Sunteți pe pagina 1din 510

# J] HOP

PROCEEDINGS OF THE TECHNICAL MEETING ON /J


mi,^ / / / / / ^

SUPER-LAMINIiR FLOW BEAillS I I S SEILS

NO

DIVISION flf SFJC'iif:]. • ' ^ ' . ' l E r . • • -' . NOLOGV

GEBniBfiTrf ;•• % ^ I

IF wc.iiM SCIE5CI A?STRA:TS

PIEMEI UilEi liiliilT IT (3H)43I3


I!
lECHIIICtt ICHMLiei liCiiPBilTEi
LITiliJEilil
L l l i i i S , EDITIi
lEIEilEi ISil
DISCLAIMER

This report was prepared as an account of work sponsored by an


agency of the United States Government. Neither the United States
Government nor any agency Thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal
liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process
disclosed, or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial product,
process, or service by trade name, trademark, manufacturer, or
otherwise does not necessarily constitute or imply its endorsement,
recommendation, or favoring by the United States Government or any
agency thereof. The views and opinions of authors expressed herein
do not necessarily state or reflect those of the United States
Government or any agency thereof.
DISCLAIMER

Portions of this document may be illegible in


electronic image products. Images are produced
from the best available original document.
LEGAL NOTICE

Neither the United Statess nor the Commission, nor any person acting on
behalf of the Commission:

A. Makes any warranty or representation^ expressed or implied^ with


respect to the accuracyj completeness 5 or usefulness of the
information contained in this report, or that the use of any
information, apparatus, method or process disclosed in this
report may not infringe privately owned rights; or

B. Assumes any liabilities with respect to the use of, or for damages
resulting from the use of any information, apparatus, method, or
process disclosed in this report.

As used in the above, "person acting on behalf of the Commission"


includes any employee or contractor of the Commission, or employee
of such contractor, to the extent that such employee or contractor
of the Commission, or employee of such contractor prepares,
disseminates 5 or provides access to, any information pursuant to
his employment or contract with the Commission, or his employment
with such contractor.
NYO 3363-6
MTI-66TR66
^ CFSTI PtICES

« G - C 5 ^ . MN , / ^
Proceedings of the Technical Meeting on:

SUPER-LAMINAR FLOW BEARINGS AND SEALS


FOR PROCESS-FLUID LUBRICATED TURBOMACHINERY

November 1 and 2, 1966


Albany, New York A550USCSMBSI

&5SIRACTS
11 S\JCXiB^- SCI^^

U. S. Atomic Energy Commission


Division of Reactor Development and Technology
Special Projects Branch
Germantown, Maryland

Prepared Under Contract AT(30-l)-3363

Mechanical Technology Incorporated


Latham, New York

E. B. Arwas, Editor

December, 1966

MTI-2349
1

k TABLE OF CONTENTS

Page
FOREWORD iii

SECTION 1 - SUMMARY 1.1


B. Sternlicht, Mechanical Technology Inc.

SECTION 2 - INTRODUCTORY ADDRESS 2.1

Review of the Role of the Atomic Energy


Commission in Process-Fluid Lubrication
N. Grossman, U. S. Atomic Energy Commission

SECTION 3 - BACKGROUND REVIEW 3.1

Process Fluid Lubrication of Turbomachinery


Bearings
-.. E. B. Arwas 5 Mechanical Technology Inc.

SECTION 4 - TECHNOLOGY I - FUNDAMENTALS 4.1

4.1 - Super-Laminar Flow in Bearings and Seals 4.1.1


C. H. T. Pan and J. H. Vohr,
Mechanical Technology Inc.
4.2 - Simulation of Turbulent Lubricant Films
in a Large Scale Apparatus 4.2.1
R. Burton, Southwest Research Institute

4.3 - Motion Picture Visualization of Laminar,


Vortex and Turbulent Flows in the Annular
Gap Between Concentric and Eccentric
Rotating Cylinders 4.3.1
J. H. Vohr, Mechanical Technology Inc.

4.4 - Conditions for the Rupture of a Lubricating


Film 4.4.1
J. C. Coyne, Bell Telephone Laboratories

SECTION 5 - TECHNOLOGY II - SEALS 5.1

5.1 - Experimental and Theoretical Study of the


Visco Seal 5.1,1
J. Zuk, L. P. Ludwig and R. L. Johnson,
National Aeronautics and Space
Administration

5.2 - The Effect of Turbulence on Visco Seal


Performance 5.2.1
^
W. K. Stair, University of Tennessee
11

TABLE OF CONTENTS (continued)

SECTION 6 - TECHNOLOGY III - BEARINGS AND MATERIALS


Page

6.1
4
6.1 - Experiments with Hydrodynamic Journal
Bearings of Various Materials and Designs
in Sodium at Temperatures to 800 F 6.1.1
F. T. Schuller, W. J. Anderson and
Z. Nemeth, National Aeronautics and
Space Administration

6.2 - Alkali Metal Bearing and Seal Development


at Space Power and Propulsion Section 6.2.1
E. Schnetzer, General Electric Co.

16.3 - Bearing and Seal Materials for Liquid Metal


Lubrication 6.3.1
S. F. Murray, Mechanical Technology Inc.

SECTION 7 - TECHNOLOGY IV - ROTOR-BEARING DYNAMICS 7.1

7.1 - Bearing Shaft System Dynamics - Mercury


Rankine Experience at TRW 7.1.1
R. Kasuba, TRW Inc.

7.2 - Calculation and Experiments on the Unbalance


Response of Flexible Rotors Supported by
Tilting Pad Bearings Operating in the
Turbulent Flow Regime 7.2.1
J. W. Lund and F. K. Orcutt, Mechanical
Technology Incorporated

SECTION 8 - APPLICATION IN LARGE TURBOMACHINES 8.1

\ 8.1 - Bearing and Seal Requirements for Liquid


Metal Cooled Reactor Systems 8.1.1
R. W. Dickinson, Atomics International

8.2 - Description of a Large Sodium Pump Concept


for Future Sodium Cooled Power Reactors 8.2.1
D. R. Nixon, Westinghouse Electric Corp.

_j8.3 - Process Fluid Lubricated Bearings for High


I
Temperature Gas Cooled Reactor Circulators 8.3.1
J. Yampolsky, General Atomics Division
of General Dynamics Corp., and D. F. Wilcock,
Mechanical Technology Incorporated

4
f
Ill

FOREWORD

The Technical Meeting on Development of Super-Laminar Flow Bearings and Seals


for Process-Fluid Lubricated Turbomachinery was held in Albany, New York on
November 1 and 2, 1966. The publication of the technical presentations made at
this meeting is sponsored by the U.S. Atomic Energy Commission and has been
prepared by Mechanical Technology Incorporated under U.S. AEC Contract AT(30-1)-3363.

The technology of lubrication in the turbulent-flow regime has practical importance


because of the recent and continuing developments of high-speed turbomachinery
lubricated with low viscosity fluids such as water, steam, and liquid metals.
Examples of such developments include compact, dynamic power-conversion turbo-
machinery for use in space, undersea, and in mobile, land-based power-plants,
as well as large coolant-flow pumps and circulators for use in nuclear reactor
installations.

The technical meeting of November 1 and 2, 1966 was held to provide for an ex-
change of Information between the various groups active in research, development
and application programs in this area. During the course of three technical
sessions, seventeen presentations were made by representatives of U.S. Government
Agencies and commercial contractors, on recent progress, current effort and anti-
cipated future requirements in process-fluid lubrication and its applications
to nuclear and aerospace turbomachinery.

This report contains all these presentations, reproduced from the manuscripts
furnished by the authors to the editor. For continuity and ease of reference,
these presentations have been assembled here according to their principal topic
into the sections listed below:

Section 1 - Summary
Section 2 - Introductory Address
Section 3 - Background Review
Section 4 - Technology I - Fundamentals
Section 5 - Technology II - Seals
Section 6 - Technology III - Bearings and Materials
Section 7 - Technology IV - Rotor-Bearing Dynamics
Section 8 - Application to Large Turbomachines

It should be noted, however, that some of the presentations included in Sections 5


through 7 covered other material, including experience with process-fluid lubri-
cation in high-speed machines.
SUMMARY

TECHNICAL METING ON DEVELOPI-ffiNT OF SUPER-LAMINAR FLOW


BEARINGS AND SEALS FOR PROCESS-FLUID LUBRICATED TURBOMACHINERY

by

Dr, Beno Sternlicht


Technical Director
Mechanical Technology Incorporated
Latham, New York

The use of process-fluid lubrication in turbomachinery presents many advantages.


It eliminates oil contamination and permits operation at pyrogenic and cryogenic
temperatures, as well as in radioactive and corrosive environments, without
lubricant breakdown. The use of process-fluid lubrication, however, requires
some fundamental changes in mechanical design. The principal new design con-
siderations include the effects of turbulent flow in bearings and seals, rotor-
bearings stability, reduced bearing damping and ultimate load capacity, oxide
deposition, selection of compatible materials and surface coatings, and others.

In the early 1960's, as a result of many high-speed machine failures, it was


recognized that the application of turbulent lubrication technology to bearing,
seal and rotor-bearing dynamics design was very limited. To our surprise, we
found that the Rumanians were leaders in this technology. As early as 1957 a
book by N. Tipei was published in which turbulent lubrication was treated. In
1963 at the University of Houston, Constantinescu presented a comprehensive
lecture on "Theory of Turbulent Lubrication." In 1965 his book, "Theory of
Lubrication in Turbulent Regime," was published. In retrospect, this technology
should have been vigorously pursued as soon as it was recognized that there was
a trend to higher speed machinery and that many of the contemplated cycle fluids
(e.g., liquid metals, water) had low kinematic viscosities which would result
in turbulent flows at relatively low speeds. In fact, turbulent and compressibl
lubrication theory and practice should have been the goals of lubrication
engineers, recognizing that laminar (Re < 1000) and incompressible (A—*-o)
lubrication analyses represent only limiting cases of the more general treat-
ment of this subject.
In recent years a number of hardware programs (e.g. , SNAP 2, SNAP 8 and SNAP 50/
SPUR) and, to a more limited extent, technology studies have been conducted in the
United States by several groups. We have organized this seminar for the purpose of
bringing together people actively engaged in this field in order to disseminate
information, present the state of knowledge and identify some of the remaining prob
lem areas. The papers were divided into the following categories;

Review
Technology
Applications

The following represents a brief summary of the meeting. More detailed discussions
will be found in the other papers which make up this report.

Present State of Knowledge

1. Visual studies have been conducted to gain better understanding of transition


from laminar to vortex and turbulent flows in annular gaps. Both concentric
and eccentric gaps were studied. The onset and development of film rupture
were also investigated visually. Theory has been developed for the boundary
conditions at film separation that are to be used with the Reynolds equation.

2. Superlaminar theory neglecting inertia effects has been developed for the
calculation of load, frictional losses, stiffness and damping for several
bearing types. This theory is in excellent agreement with practice.

3. Several dynamic seal configurations have been investigated experimentally.


These include screw seals (visco seals), rotating channels, slingers, and
squeeze seals. The experimental studies revealed a fundamental difference
in gas ingestion mechanism between grooves on the shaft and grooves on the
housing.

4. Stability of several bearing types has been investigated experimentally.


Considerable duplication of effort appears in this area. The bearings studied
included plain cylindrical, two, three and four axial groove, pressure pad,
three lobe, hybrid, herringbone, floating ring, pivoted shoe (tilting pad),
etc. Tilting pad bearings exhibited greatest stabilityi herringbone and
floating sleeve bearings were next best.
5, Theory is available for obtaining stiffness and damping coefficients for
several bearing types operating in turbulent regime. These coefficients
can be coupled in the bearing-rotor dynamics analysis to obtain rotor
response. The agreement between theory and practice for critical speeds,
amplitudes of vibration and threshold of instability is good.

6, Several machines have been successfully developed using process-fluid


lubricated bearings and seals, (Due to the lack of technology information,
however, too much trial and error was involved in these developments.)

Remaining Problem Areas

1. Turbulent lubrication theory, incorporating inertia effects, is required,

2. Theoretical analysis of hybrid journal and thrust bearings^ also incorpora-


ting inertia effects^ is required.

3. Theory is required for visco-seals and other potential dynamic seals.

4. Reliable rugged instrumentation for dynamic measurement of bearing and seal


film thickness operating in high temperature liquid metal environment is
very badly needed for both experimental work and monitoring of machinery.

5. Bearing and seal materials require immediate attention. Cemented carbides^


which are good from the standpoints of sliding behavior, corrosion resistance
and high-temperature strength, present major design problems due to their
low coefficients of expansion, shock resistance, manufacturing difficulties
and high cost. Consideration should be given to other materials and to
coatings,

6. Theoretical and experimental studies are required on cavitation in bearings


and seals operating in liquid metal environment. The complex flow regimes
that take place in bearings and seals must be considered in such a study as
opposed to simple cavitation in ventury, plate, disk and other experimental
apparatus.

7. In the case of hybrid (externally pressurized with hydrodynamic effects)


bearings, erosion of restrictors and pockets must be investigated in order
to ensure long life, trouble-free operation. This is basically a fluid
mechanics and materials problem.
8, Considerable attention must be paid to the design and manufacturing of the
large sodium cooled reactor bearings and seals. Because of their size they
present major problems of material choice, method of attachment (of sleeves,
runners, bearings), flow requirements, power loss, stiffness, damping, etc.

9, Testing in liquid metals is very expensive and time consuming; therefore, it


must be used in conjunction with theory in order to minimize the number of
tests and establish broadly applicable technology. Accurate, well-instrumented
tests are therefore essential (such instrumentation has not been proven out
at this time). Theory must be available for guidance and correlation (con-
siderable theory is still missing (refer to items 1, 2, 3 and 6 above).

The above-mentioned problem areas require immediate attention if reliable process-


fluid turbomachinery is to be developed. It is hoped that technical papers on
these subjects will be presented in the future and that other similar meetings
will be held in order to exchange information. The papers and meetings will be
most valuable if the investigators clearly present the data, accuracy of data, and
point out problem areas. It is further hoped that such meetings will help to
prompt complementary efforts and to minimize duplication.
INTRODUCTORY ADDRESS

y. Review of the Role of the hi,S..


Atomic Energy Commission
in Process-Fluid Lubrication

by

Nicholas Grossman
Chief, Special Technology Branch
Division of Reactor Development and Technology
USAEC

I want to thank Mechanical Technology Incorporated for making this en-


gineering get-together possible. I consider this meeting an important milestone
in the technology of process-fluid lubrication; it heralds the emergence of
process-fluid lubrication from the laboratory and the test bench to practical
engineering design. Naturally, we all realize that our analytical knowledge
in this field is indeed very small compared with the vast areas yet to be ex-
plored -- but the successful demonstration of process-fluid lubrication to
rotating machinery is here -- with us today.

Perhaps it will be useful to describe -- in bare outline -- the goals of


the Atomic Energy Commission's program in power reactor development; our
interest in process-fluid lubrication and the specific needs for process-fluid
lubrication.

Many of you are aware of the long standing, well organized and highly
successful program in gas lubrication coordinated by the Fluid Mechanics Branch
of the Office of Naval Research. Each of us here today is a direct beneficiary
of the farsighted and effectively administered ONR program. There are a number
of enthusiastic supporters who materially contributed to this undertaking, but I
feel we should specifically mention two of our friends, whose unfailing devotion
stands out as an example for us all: Mr, Stanley Doroff of ONR, who is directly
responsible for procuring the support,defending the program, and dispensing the
funds with Solomon-like impartiality and wisdom -- and Professor Dudley Fuller
of Columbia University, who in the role of Technical Secretary of the Coordinating
group, has been the program chairman, recording secretary, conciliator and tower
of strength.

I recall that during the 1965 summer meeting of the ONR Gas Bearing Co-
ordinating Group, Columbia University was our obliging host. We were fortunate
to have Dean Robert Dunning as our keynote speaker. As you know. Dean Dunning
was one of the pioneers who helped move atomic energy from the physics laboratory
into the realm of engineering.

During his talk to the ONR Gas Bearing Group in 1965, Dean Dunning mentioned
how he and his engineering associates working on the design of the gaseous
diffusion plant some twenty years earlier, recognized the practical advantages
of gas lubrication and made record notes accordingly! Thus if we wish to
establish a date to connect gas lubrication and atomic energy, we can go back to
the wartime Manhattan Engineering District and state that engineers working with
atomic energy on an industrial scale recognized the natural affinity between
process-fluid lubrication and atomic energy at the very outset. In my opinion,
had rigid security classification not been necessary during the war, the develop-
ment of process-fluid lubrication would have progressed at a much faster rate,
and certainly would have been a practical industry application at a much earlier
date.

It is well known that the idea of gas-lubricated machinery was not incorp-
orated in the gaseous diffusion plant built during the war. It is at this
junction that we must appreciate the foresight of the Fluid Mechanics Branch of
ONR for recognizing the industrial potential of gas lubrication and organizing
a sound research program to cover this vast field in an orderly way. Naturally,
a clever idea like gas lubrication cannot be kept as a monopoly of any one group,
and just as it was recognized in America, similar development efforts were pursued
in Europe.

One of the obvious consequences of the war in Europe was the acute shortage
of modern central electric power stations. Consequently, there was greater
impetus to develop central power stations -- and their construction was pushed
much more vigorously in Europe than in the United States. It is understandable.
therefore, that industrial applications of gas bearings in gas blowers appeared
in Europe a few years ago.

Coming back to the Atomic Energy Commission -- its role in the promotion
of industrial application is specified in Public Law 83-703, popularly known as
"The Atomic Energy Act of 1954." Section 3 states in part: "It is the purpose
of this Act to effectuate ....... a program of conducting, assisting, and
fostering research and development in order to encourage maximum scientific and
industrial progress." How successful the Commission has been in carrying out its
charter can be readily assessed in terms of nuclear power stations built or being
constructed, the use of radioisotopes and other byproducts now in common use,
and nuclear propulsion for naval purposes. What is amazing about the success
of atomic energy as an established segment of American industry is not that it
came into existence, but rather the phenomenal speed of that development which
surprised even its most optimistic advocates.

Looking at the typical American commercially built nuclear fueled central


power stations, it may appear that -- after stripping away some superficial
differences -- it is basically a closed-cycle steam power plant "burning" uranium
as its heat source. Looking at the typical British nuclear power station, it
appears as a closed gas cycle heat generator, "burning" uranium as its fuel. Both
types have proven reliability and economic value.

Important components of these systems are circulators, in many cases using


process-fluid lubrication. European gas-cooled reactors -- such as Dragon --
use gas-bearing motor-driven circulators. Just two years ago a significant
milestone was achieved when Mil, under an AEC- Bureau of Mines contract, demon-
strated for the first time a gas bearing turbocirculator operating in a closed-
loop Braytoa cycle. Sipce that time a great deal of technical information has
been obtained from that program.

Today you will hear about turbine-driven circulators and power conversion
systems using water and liquid metal bearings and seals in a turbuluent regime.
Future developments will lead to the engineering application of two-phase
lubricated systems in turbomachinery Since the AEG is developing two-phase
systems, the Commission's interest in process-fluid lubrication is therefore
apparent. In feet, most of the advanced process-fluid lubricated machines
2.4

have been developed under AEC and Navy contracts. d

What then is the goal of the AEC for future development? I will restrict
my remarks to the program of the Division of Reactor Development and Technology.
Current commercial nuclear power plants use a uranium "burning" reactor operating
in the thermal neutron spectrum which converts a portion of the uranium into a new
fissionable material: plutonium. There is thus a newly created material that can
be used for power generation. However, the amount of new material produced is
small compared with the uranium that was used or "burned" in the fission process.
Therefore, looking at this process from the point of view of most effective
utilization and conservation of our national reserves, thermal reactors are con-
sumers of fissionable material. One of the amazing aspects of nuclear fission --
at least to us engineers -- is that in a suitable fast neutron spectrum more new
fissionable material can be "created" than is being "used". This concept is
popularly known as "breeder reactor". The orderly development of this system
is a high priority goal of civilian nuclear power development programs.

The required combination of nuclear physics and engineering for nuclear


reactors operating in the fast neutron spectrum points to the alkaline liquid
metals as the heat transport medium. This is the concept that has been carried
furthest, and the one that has shown to be technically feasible in actual reactors,
both in the United States and Europe. Therefore, it is logical that the Division
of Reactor Development and Technology actively pursue an orderly program leading
to a reliable alkaline liquid metal bearing and seal system for rotating machinery.
I have not seen the reports and papers to be presented during this s3miposium ahead
of time, but I am certain the authors will present the technical details and
enumerate the potential advantages and benefits of liquid metal bearings.

Before closing I want to state that the Space Nuclear Systems Division of the
AEC, as well as other government agencies, are also pursuing the development of
liquid metal bearings and seals, where their application offers attractive benefits
to their specific requirements. It is one of the purposes of this s5nnposium to
afford us all an opportunity to become better acquainted with the efforts of
other groups, and offer the benefit of our experience -- to one another. I am
very pleased with the progress to date and I hope you will share this pride.

Thank you.
3.1

BACKGROUND REVIEW
'-'11
V ' PROCESS-FLUID LUBRICATION OF TURBOMACHINERY BEARINGS

by

E. B. Arwas
Mechanical Technology Inc.
Latham, N. Y.

ABSTRACT

Process-fluid lubrication is being increasingly specified for closed cycle, high


speed and high temperature (as well as cryogenic) turbomachinery applications.
This is being done to preclude oil contamination of critical system components,
to permit bearing operation at the system temperatures, to minimize sealing re-
quirements and for other similar reasons. The use of process-fluid lubrication,
however, requires some fundamental changes in mechanical design to cope with the
limitations of many cycle fluids when used as lubricants. These design consid-
erations include the effects of turbulent flows in the bearing gap, bearing
stability, reduced bearing damping and ultimate load capacity, oxide deposition,
selection of compatible bearing materials and surface coatings, potential pivot
fretting and others. XD_iJils__pa^iaX4 the advantages, problem areas, and design
approaches that have been used in current process-fluid lubricated machines a*e-^-^
are discussed. ^il//^

INTRODUCTION

The theme of this technical meeting is process-fluid lubrication and is attendant


phenomenon in high speed applications, fluid-film turbulence. Process-fluid
lubrication is not a novel concept - in fact, it predates conventional oil
lubrication. The marked difference today, however, lies in its application in
the sophisticated turbomachinery that is being developed by the nuclear and
aerospace industries, where the bearing and seal requirements are very
exacting and the margin for error is small. The development of bearings and
seals and the selection of rotor configuration by means of trial and error can
be extremely expensive in these advanced applications. It may be minimized, even
altogether eliminated in most cases, by rational design based on a thorough
understanding of the technology of lubrication with low viscosity fluids.
3.2

The papers presented in this technical meeting have ranged over much of the recent
^
work as well as the current and future needs of this technology. Thus, one of the
presentations (Section 8.1) discussed the bearing and seal requirements for projected,
large, nuclear, liquid sodium facilites. Other presentations (Sections 6.2, 7.1,
and 8.3) reviewed recent and current turbomachinery developments involving water
or liquid metal lubricated bearings and seals operating in the turbulent flow
regime. In still other presentations (Sections 4 through 6 ) , some of the most
recent theoretical approaches, empiricisms and experimental analyses of turbulent
flow lubrication with low viscosity fluids are described. The important point is
that all this effort, comprising research, practical hardware development and
definition of specification for future applications, is proceeding simultaneously.
This is a healthy condition for a technology. It is the hope behind technical
meetings such as this one to provide forums where the hardware development engineers
and the research investigators are kept cognizant of one another's progress and
changing requirements

We at M.T.I, have also tried to achieve this balance of theoretical investigation


and practical application. Under a number of programs conducted for the Atomic
Energy Commission, for the National Aeronautics and Space Administration and for
the Agencies of the Department of Defence, we have sought to:

(a) develop a rational theory of turbulent lubrication, relying only


on accepted precepts of turbulent flows generally and without recourse
to new constants or "correction factors" to provide agreement with
the relatively sm.all amount of turbulent bearing test data available
in the open literature.

(b) prepare numerical procedures and computer programs for calculating the
steady-state and dynamic characteristics of various bearing geometries,
from the turbulent lubrication theory,

(c) test the validity of the theory by means of bearing and rotor-bearing
dynamics tests with different bearing geometries and rotor arrangements
and,

(d) apply the verified theoretical analysis and the computation procedures
in the design of bearings and seals for hardware developments.
3.3

The theoretical analysis and its experimental verification have been published in
contract reports and in the open literature (References 1 through 5). The results
obtained to date and the questions still remaining are noted in Section 4.1,
together with our current conclusions regarding the regions of bearing operation
where vortex flows, developed turbulence and fluid inertia effects either singly
or in combination, will govern performance. The information gathered in this
research is being applied to practical bearing designs, such as the ones for
HTGR circulator, which are described in Section 8.3.

One aspect of turbulent flow lubrication that appeared to us to be particularly


important is the sharp rise in power loss with speed, that occurs immediately
after the point of transition from laminar flow. We believe that the attention
paid to this problem will increase as experience is gained with lightly loaded,
high surface-speed bearings of turbomachines lubricated with water, liquid metals
and other low viscosity fluids. We also believe that this problem as well as
other factors such as compactness, weight-saving and general design simplifications
will lead to a greater acceptance of flexible rotors with small shaft sizes and
greatly reduced bearing and seal losses. In high speed bearings design, parti-
cularly when flexible rotors and supported, the dynamic characteristics of the
bearings and of the coupled rotor-bearings system are of primary importance.
Accordingly, we have laid great stress in our work to date on accurate determina-
tion of the stiffness and damping characterisitcs of bearing films. In Section 7.2
the results of an experimental program in which three different rotor arrangements
were operated on turbulent flow bearings up to or through the third system critical
speed in each case, are described. The lubricant used in this instance was a
silicone fluid whose viscosity is intermediate between that of water and those of
sodium and potassium.

Currently, under continuing AEC sponsorship this work on bearing and rotor-bearing
dynamic performance is being extended to operation with a liquid metal lubricant
(NaK). This test program is currently underway, the facility and the proposed
test program are briefly described in an Appendix to this paper.

Since the specifics of the effort at M.T.I, are covered here in the presentation
of other M.T.I, engineers, the balance of this presentation will be limited to a
general review of:
3.4

1. some of the advantages of simplification (and, hence, reliability) and


improved turbomachinery efficiency that are made inherently possible
when process lubrication is used, and

2. some of the important design considerations in lubeication of high


speed machines with fluids that have low viscosity and poor
lubricityJ

ADVANTAGES AND PROBLEMS OF PROCESS FLUID LUBRICATION


It is appropriate here, to briefly review the factors that we take for granted
when using hydrocarbon oils as lubricants as well as the problems that weigh against
their use in some closed cycle, high temperature systems. These are listed in
Table 1 below,

TABLE 1
CHARACTERISTIC PROPERTIES OF HIGH SPEED. OIL LUBRICATED
SLIDER BEARINGS

ADVANTAGES:

High ultimate load capacity


Excellent boundary lubrication
High damping
High instability threshold
Non-corrosive
Low surface tension
Good wettability

LIMITATIONS IN CLOSED CYCLE. HIGH-TEMPERATURE SYSTEMS:

Contamination of System components


Complex cycle-fluid to oil sealing
Limited operating temperature range
High friction
Oil breakdown in radio-active environment
Complex auxiliary system
Potential fire hazard
3.5

While load capacity is not generally a problem in high speed machines, the
relatively high viscosities of many oils, combined with their excellent boundary
lubrication properties offer a very desirable margin of safety in event of mal-
function or abnormal conditions. In event of a failure, such as the loss of
blades in a turbine and the consequent very high dynamic loads, the high ultimate
load capacity and the boundary lubrication properties may, in some cases,
reduce the severity of the damage to the rotor.

High damping is clearly useful in limiting vibration amplitudes at system re-


sonances. Advantage can also sometimes be taken of the high damping to suppress
instabilities or to control the amplitudes of whirl.

Oils are generally non-corrosive, which increases the choice of bearing materials.
Finally, the low surface tension and good wettability improve the start up and
we believe, on the basis of some fairly preliminary investigations (References 6
and 7) , that these properties also tend to reduce the likelihood of cavitation
erosion in most applications. This is because they tend to produce a steady
striation type of film rupture in the regions of negative static pressure in the
film, instead of small, gas or vapor filled bubbles which can then collapse as they
move to positive static pressure regions.

Despite these advantages,and they are important ones, oil lubrication is difficult
in many of the advanced, closed cycle systems required in nuclear, aerospace and
undersea application. Thus, oil fouling and contamination of some of these closed
systems cannot be tolerated and it is extremely difficult to prevent if oil
lubricated bearings are used. The problem in the way of development of dynamic,
zero-leakage seals are, we believe, far greater than those of process-fluid
lubrication.

Lubricating oils are also limited in their temperature capability and cannot be
used at temperatures exceeding say 600 to 700 F, or at very low temperatures
approaching the cryogenic range. Thus, complex bearing temperature control provisions
are needed, as well as use of heat dams or other undesirable design comprises.

Because of high viscosity, friction losses in high speed oil lubricated slider
bearings are large and represent a sizeable penalty on efficiency.
Hydrocarbon oils cannot sustain radio-activity for long periods without breaking
down, which rules them out in some nuclear applications.

The complexity of the auxiliary lube-oil system cannot be minimized. The complex
of lube pumps, coolers, filters, controls and other items is complicated and
bulky. Moreover, while its component parts are generally well developed they have
nonetheless proved to be a major source of unreliability. Finally, of course, the
lube-oil system sometimes constitutes a fire hazard.

Turning now to process-fluid lubricated systems, we can make a similar, general


assessment of advantages and potential problem areas. The advantages are listed
in Table 2 below. Thus, the elimination of the auxiliary lube system removes the
problem of oil fouling or contamination of critical system components, as well
as the need for cycle-fluid to lube oil seals.

TABLE 2
ADVANTAGES OF FULL-FILM PROCESS-FLUID LUBRICATION
IN HIGH SPEED. HIGH TEMPERATURE.
CLOSED CYCLE SYSTEMS

1. ELIMINATION OF AUXILIARY LUBE SYSTEM:


No oil fouling or contamination
No oil to cycle-fluid seals
Removes oil fire hazard

2. DESIGN SIMPLICITY AND IMPROVED EFFICIENCY


OF TURBOMACHINERY
Reduced machine length
Reduced volume and weight
Bearings operate at system temperatures
Bearings operate at system's level of radio-activity
Low bearing friction (very low with gases and vapors)

Substantial improvements in design simplification and increased efficiency of


the turbomachinery are made possible by:

1. reduction of axial length due to the elimination of the cycle


fluid to oil seals
3.

2. reduction in volume and weight due to the elimination of the lube-oil


system

3, the bearings have at least the potential of operating at the local


rotor temperatures, thus eliminating the need for heat shields
and the problem of thermal gradients and"their associated distortions

4, the bearings can operate at the level of radio activity of the system

5. the bearing friction losses are low, particularly if the lubricant


is a gas or vapor.

Conversely, of course, in applying process fluid lubrication, careful attention


has to be paid to a number of factors, many of which will depend on the particular
process fluid being used as the lubricant. These factors are noted in Table 3
on page 3.8

Thus 9 in the case of gas or vapor lubrication, the self-generated bearing load
capacity is extremely small (of the order of a few PSI), due to the very low
absolute viscosities of gases and vapors. Accordingly, externally pressurized
bearings have to be used to support large steady state or dynamic loads.
The fluid film damping in gas or vapor lubricated bearings is relatively small,
and accurate rotor response analysis should be made to insure that system and
component resonant frequencies are outside the operating range. The compress-
ibility of gases and vapors complicates the analysis of gas bearings, introducing
an additional parameter (the compressibility number A) and making the lubrication
differential equation non-linear. An important practical consequence of the
non-linearity is that the stiffness and damping of gas-bearing films are highly
frequency dependent. In particular, gas bearing damping tends to diminish rapidly
when the film is subjected to high frequency excitations. Since excitations at
multiples of the running speed are not unusual in some turbomachines (e.g. alternator
rotors), they can produce large vibrations if they excite a system resonance,
due to the reduced damping of the gas film. It is thus particularly important to
analyse accurately the dynamic behavior of high speed, gas bearing supported
systems,

In the case of externally pressurized bearings, gas compressibility also causes


pneumatic hammer. This is prevented by avoiding recessed volumes in the film, by
TABLE 3
SOME DESIGN CONSIDERATIONS WITH PROCESS FLUID LUBRICATION OF HIGH SPEED TURBOmCHINERY

SELF-ACTING BEARINGS EXTERNALLY PRESSURIZED BEARINGS

Very low l o a d capacity Provision of pressurized gas


Low damping Low damping at high speeds
Compressibility effects: Gas compressibility effects:
analytical complexity analytical complexity
frequency dependent K and B frequency dependent K and B
reduced instability threshold fractional frequency whirl
Thermal distortions pneumatic hammer
S t a r t - s t o p wear No boundary lubrication
No boundary lubrication

Vapor As above plus: As above plus:


(wet or with
low superheat) Condensation Condensation
Bearing erosion Bearing erosion
Restrictor erosion

Liquid Low to moderate ultimate Provision of pressurized liquid


load capacity Turbulence
Moderate to poor damping Moderate to poor boundary lubrication
Turbulence: Corrosiveness (liquid metals and some
high friction cycle fluids)
large attitude angle Oxide deposition
Moderate to poor boundary Surface reactions
lubrication Restrictor and bearing erosion
Corrosiveness (liquid metals
and some other cycle fluids)
Oxide deposition
Surface reactions
High surface tension
Poor wettability
Cavitation erosion
3.9

using inherently compensated restrictors and, where practical, by small external


damping provisions.

The low load capacity ®f self-acting gas bearings makes them relatively intolerant
of thermal distortions, which can adversely affect the film geometry or produce
misaligning moments and edge loads. This can be aggravated by the fact that the
thermal capacity of the gas flow through self-acting bearings is small, so that
the heat generated in the bearing is removed by conduction through the bearing
elements, resulting in thermal gradients and corresponding distortions. In past
applications J the problem of thermal distortions has been controlled by the use
of heat dams and heat shunts, by locating the bearings in isothermal regions of
the rotor and by external cooling. Currently, effort is being devoted to develop-
ment of flexure mounts and conformable surfaces to permit bearing operation in the
presence of large thermal gradients.

Gases provide little or no boundary lubrication so that in all cases (in parti-
cularly for self-acting gas bearings which do not use hydrostatic jacking and
where a large number of starts and stops are anticipated over the life of the
machine) the materials must be selected to sustain frequent rubs without damage.

Currently, experience with vapor lubrication is much smaller than that with gas
lubrication. However, since lubrication of the bearings located at the turbine
end of a rotor with the vapor supplied to the turbine (e.g. steam, metal vapors
or others) can potentially result in major simplification and increased relia-
bility of the turbomachine5 an increased level of effort in this area of
lubrication technology is anticipated. Here, the effects of phase change, bearing
erosion and, in the case of externally pressurized bearings, restrictor erosion
have to be considered (Ref. 6) .

With liquid lubricants, the problems, at least from the standpoint of load
capacity, fluid film damping and stability, are somewhat less critical than
with gas lubrication, due to the higher absolute viscosities or liquids and
the absence or compressibility effects.

Many of these liquids, however, have very low kinematic viscosities, leading to
onset of superlaminar flows in high speed bearing films. Table 4 on Page 3.10
TABLE 4 -- VISCOSITIES OF SOME PROCESS FLUIDS

ABSOLUTE KINEMATIC
ABSOLUTE VISCOSITY, n VISCOSITY KINEMATIC VISCOSITY,v VISCOSITY
FLUID (Ib-sec/in. ) RATIO* ...__{ in,, i/secl RATIO*

SAE 10 oil (120F) 35 X 10 -7 1 44 X lO"^ 1

Water (120 F) 0.812 X 10 1/43 0.890 X io~3 1/49

Gasoline (120 F) 0.725 X 10 1/48 1.035 X 10-3 1/33


-7
NaK 78 (400 F) 0.536 X 10 1/65 0.695 X 10-3 1/63
-7
Mercury (400 F) 1.510 X 10 1/23 0.123 X 10-3 1/358
-7
Air (70 F, 15 psia) 0.026 X 10 1/1340 22.7 X 10-3 1/2

Steam (600 psia, sat.) 0.027 X 10 1/1290 1.39 X 10-3 1/32

*Ratio of fluid viscosity to viscosity of SAE 10 oil.


3

lists the absolute and kinematic viscosities of a number of fluids and compares
them with those of an SAE 10 oil. This illustrates the relative load capacities
and the turbulence onset speeds, since the absolute viscosity of the lubricant
is a measure of the potential bearing load carrying capability (for self-acting
bearings), while the speed at onset of turbulence is directly proportional to
the kinematic viscosity of the fluid.

The onset of turbulence in a bearing is important, because:

a) it is accompanied by a very sharp rise in power loss

b) it results in increased load capacity and changes the dynamic character-


istics (i.e. the stiffness and damping) of the bearing

c) it produces a moderate increase in attitude angle with, potentially, a


greater tendency to instability.

All of these changes are important in high speed bearings so that, for rational
bearing and rotor-bearing dynamics design, it is necessary to know whether the
bearing film is laminar or turbulent and when fluid inertia affects are sizeable.
The appropriate theory should then be used to compute the steady-state and dynamic
performance characteristics of the bearing.

Some of the other important considerations in designing process-fluid lubricated


bearings, where the fluid is a low viscosity liquid such as water or a liquid
metal are also noted in Table 3. The bearing surface materials have to be
selected to prevent corrosion and surface reactions. Oxide depositions in the
bearing and seal regions are prevented by insuring that these elements do not
constitute cold traps. Finally, cavitation damage believed to be more likely to
occur with fluids that have high surface tension, and poor wettability. This may
be prevented, or at least minimized by pressurization to prevent film rupture
and by suitable materials selection.

Process Fluid Lubricated Turbomachinery

A recent study conducted jointly by the Army Engineers Reactor Group and MTI
(Ref. 7 ) , illustrated some of the potential advantages of process fluids
lubrication. In this study, the oil lubrication systems and the bearing power
losses of three closed loop gas turbines developed in connection with the U.S. Army
program on compact, mobile nuclear power conversion systems were studied. They
were then compared with a gas lubricated system (CSG-1) that has been proposed and
for which a dynamic simulator was built and evaluated on externally pressurized
gas bearings. Table 5, (Page3.13) reproduced from Ref. 7, lists the power losses
in each case. It is noted that here even for machines in the 2000 shaft H.P.
range, there is a significant reduction in power loss associated with process-
fluid lubrication. With machines having a smaller power output, the percentage
reduction in power loss will be still greater.

The simplification of the lubrication system and controls is seen in Figures 1


and 2 (also reproduced from Ref. 7), which show respectively the oil lubrication
system of the CSN2 unit and the process-fluid lubricated system of the CSGl. Note
that in this latter case, the pressurized gas for the bearings is bled off from the
compressor discharge. Figure 3 in a photograph of the CSN2 lube oil distribution
panel and lube oil skid.

The dynamic simulator of the CSGl was built and tested at MTI The rotor weight
and design point speed of the simulator were 90 lbs. and 28,000 RPM respectively.
The unit was operated at up to 30,000 RPM. Figure 4 is a photograph of the simulate
parts. It should beroted, however, that full development and operational experience
with gas lubricated, closed loop gas turbines in the 2000 HP range has not been
undertaken, so that the potential long term operation of such units and their
tolerance to off-design conditions are not yet known.

To date, the practical experience with gas bearing turbomachinery has been princi-
pally with motor driven gas bearing compressors used mostly in nuclear and chemo-
nuclear installations. There are a number of such units operating in the United
States and in Europe (Ref. 8). The several machines operating in the United States
have accumulated over 40,000 hours of highly successful field operation, the
longest time on any one machine being 28,000 hours.

For fewer turbine driven gas bearing machines have been developed todate. Figure 5
shows the first such unit to have been developed and tested at its design speed
(24,000 RPM) and its design point turbine inlet temperature (1300 F). This unit
was developed and operated for the U.S. Bureau of Mines and the U.S. Atomic Energy
TABLE 5
COMPARISON OF BEARING SYSTEM PARASITIC LOSSES FOR OIL- AND GAS-LUBRICATED TURBOCOMPRESSORS

Oil-lubricated Machines Gas-lubricated Machines


Machine designation TCS 560B TCS 670-2 CSN-2 CSG-1 Bu Mines/AEC
and rated speed 18,000rpm 18,338rpm 22,000rpm 28,000rpm 24,000rpm

Rated turbine shaft 1,275.0 2,235.0 2,190.0 2,235.0 84.8


power - KW

Bearing lubricant viscous 28.1 0.91 0.69


shear losses - KW •k
SS 7
Additional heat transfer 8.9 - „

losses to the lubricant-KW


* - __
Compressor bleed losses 13.4 3.89 5.0
for buffer-gas flow-KW
* *
Compressor bleed losses - _ - 0.89 0.15
for gas-bearing pressuri-
zation - KW

Compressor bleed losses - _ - 0.46


for gas-bearing cooling-KW

Power for auxiliary pumps, 2.86 2.61 11.9 - -


fans, coolers, etc. - KW -

Total bearing system losses 4.177o 3.02% 3.25% 0.08% 1.53%


as a percentage of rated
turbine power

Notes: 1. Numbers marked with an asterisk are measured values. All other numbers are calculated
values.
2. Bearing loss data for the CSG-1 machine is based on the CSG-1 rotor-bearing simulator
described in Reference 2.
3. The above losses are for the turbocompressor bearing system only. Losses associated
with reduction gearing and generator bearings are not included.
HI I-2
3.14

Commission by MTI, This unit, which has a rated shaft turbine power of 84.8 HP,
is representative (in terms of gas circulation alone) of 100 to 300 HP circulators.
In terms of closed Brayton Cycle power generation, the turbocompressor is re-
presentative of the gas generator section for a 30 to 100 KW(e) gas-turbine power
plant. Over 125 hours of test operation have been accumulated with this machine
at turbine inlet temperatures ranging from 1000 to 1400 F.

Gas bearing supported compressors and expanders are also required for cryogenic
systems where very long, maintenance free operating life is required. Recently a
two stage regenerative helium compressor was operated at MTI at speeds up to
114,000 RPM.

This particular unit, shown in Figure 6, has a rotor weighing 1.5 lbs. which
comprises the two regenerative compressor stages, an internal cooling fan and
thrust plate and a 3 phase synchronous, hysteresis type drive motor. The rotor
is supported in two 3/4" diameter tilting pad bearings, designed for stable
operation at up to 150,000 RPM. The unit is cited here, because in the course of
developing it, some preliminary experimental investigations were made of the
influence of electro-magnetic forces on such small, high speed self-acting gas
bearings. When a clean, sinusoidal AC voltage was applied, the unit operated at
up to 114,000 RPM without difficulty. When, however, an approximately square
wave output was used, the system harmonics resulted in large, unbalanced magnetic
forces on the bearings which exceeded their capabilities. Figure 7 illustrates
orbits obtained in the two cases. These show the need for careful electrical
design, to prevent large, unbalanced magnetic forces in designing motors and
alternators which are to be supported in self-acting gas bearings.

Water lubricated bearings have been used extensively and successfully in many
applications. Water lubricated pumps are, for example, commonplace in marine
and other service. Water lubricated canned-rotor pumpsj using graphitar bearings
are standard in nuclear submarine service.

Water lubricated bearings are currently being developed for the HTGR helium
circulators, because of their advantages in this application^ as discussed in
Section 8.3 of this report. Figure 8 is a drawing of the test rig for this
machine.
3.15

Liquid metals have also been used as bearing lubricants. Thus, mercury is the
lubricant used in the SNAP 2 rotating unit discussed in Section 7.1 of this
report and in other,earlier units (Ref. 9 ) .

Potassium lubricated bearings are contemplated for the potassium-Rankine cycle


turbomachines being developed for space power. Much useful technological informa-
tion on bearing materials for alkali metal lubrication, as well as some testing
of alkali metal lubricated bearings has been conducted in connection with SNAP
50/SPURand other ^programs (Refs. 10, 11 and 12).

An example of successful application of process-fluid lubrication using the alkali


metal process-fluid is in the Sodium-Potassium pump motor assemblies for the
SNAP 8 system (Ref. 13). Here, process-fluid lubrication permitted achievement
of hermetically sealed units that incorporate on a single shaft the pump impeller,
a 400 cycle, three phase induction motor and an internal lubricant coolant cir-
culating pump, supported on NaK lubricated journal and thrust bearings. Both the
journal and thrust bearings in this instance were of the tilting pad type for
reasons of stability and self-alignment. Several of these units have been operated,
accumulating over 4000 hours of test time. One unit has operated for over 3000
hours. Reference 13 notes that the results of the test program of the process
fluid lubricated SNAP 8 NaK pump-motor assembly has been most satisfactory

In conclusion then, process-fluid lubrication offers important advantages in


simplification, reliability and efficiency which have prompted its acceptance
for turbomachinery application in the nuclear and aerospace industries. The use
of low viscosity fluids with poor lubricity, however, poses a number of problems
some of which were noted above. In order to cope with these problem areas and
to minimize the need for trial and error, the bearing and rotor-bearing dynamics
design should take account of the results of the technological studies of lubri-
cation with low viscosity fluids. In the case of gas lubrication compressible
flow analysis should be used. In the case of liquid lubricated bearings, the
Reynolds Number for the fluid film and the clearance ratio of the bearing will
establish whether laminar or turbulent flow analysis should be used and whether
fluid inertia terms are important. Materials selection is also extremely critical
in view of the poor lubricity, the corrosive nature of many of the fluids and the
susceptibility to cavitation damage in some instances. New, electrical aerodynami
and mechanical design approaches are also necessary to reduce bearing loads and
otherwise compensate for the lower performance limits of bearings lubricated with
low viscosity fluids. Table 6 (Page 3.17), for example, shows some of these de-
sign considerations for machinery supported in self-acting gas bearings.

To date, a numer of high speed process-fluid lubricated machines using gases,


water, liquid metals and other fluids as lubricants have been successfully
developed and used. Some of these were also noted above. Vapor lubricated bear-
ings, where phase changes may occur in the bearing or its restrictor system, have
not yet been applied, partly because of inadequate technology in this area. How-
ever, since vapor lubrication offers important advantages in some instances, such
as for rotors supporting vapor turbines, more technological activity in this
area is required.
3.17

TABLE 6
SOME DESIGN CONSIDERATIONS WITH SELF-ACTING GAS BEARING MACHINES

STABILITY - Highly stable bearings, e.g. tilting pad


and herringbone grooved bearings
Preloading
Clearance selection

THERMAL DISTORTIONS - Steady-state and transient thermal analysis:


Minimize thermal gradients
Location of heat shields
Coolant flow required
Flexible mount of bearings
Make bearings support structure independent of housing

LOW LOAD CAPACITY - Aerodynamic balancing: impulse turbines, pressure


balancing holes, labyrinths, scalloping compressor
end plates, etc.
Electrical force balancing: 4 pole machines, rotor
to stator concentricity, phase balance, etc.
Hollow shafts
(External pressurization should be used if
above factors are insufficient to provide
required load capacity)

LOW DAMPING - In-place balancing


Rotor response analysis:
Rotor operation away from rigid body critical
Speeds and below flexural critical speed
Avoid locating bearings at nodes
Avoid mass shifts:
Rigid rotor
Single piece construction
High shrink fits
Pre-compression of motor laminations
APPENDIX

E. B. Arwas and W. D. Waldron


Mechanical Technology Inc.
Latham, New York

Under AEC Contract AT(30-1)-3363, MTI is conducting an investigation of lubri-


cation with liquid metals and other low viscosity fluids.

Earlier tasks included:

a) development of turbulent lubrication theory including the coupled


effects of shear and pressure gradient flows (Ref. 2)

b) exploratory, primarily visual study of film rupture in bearings lubri-


cated with a low viscoscity liquid (Ref. 14)

c) analytical and experimental investigation of flexible rotor operation


on tilting pad bearings operating in turbulent flow regime (Ref. 15
and paper 7.2 of this report).

As part of the current task, tables of the steady-state and dynamic characteristics
of a number of tilting pad bearings (four-pad bearings with L/D = 1/4, 1/2 and 1
and six-pad bearing with L/D - 1/4) have been computed for Reynolds numbers up to
60,000.

In the parallel, experimental work under the task currently in progress, the
steady-state and dynamic load capacity, as well as the stiffness and damping
coefficients are being accurately determined for a four pad, L/D = 1 bearing,
for comparison with the theoretical data. The lubricant used is NaK 78 and
the measurements made are of the journal location and orbit size and geometry,
for a range of values of steady-state loads, dynamic loads and rotational speed.
The characteristics of the test apparatus are:

Rotor weight - 57 lbs.


Speed range - 3000 to 28,000 RPM, infinitely variable (design
speed = 24,000 RPM) .
Bearing size - 2" dia x 2" long
Bearing type - four pad, tilting pad with 0.55 pivot position
and 0.5 geometrical preload factor.
3.19

Machined clearance ratio - 0.004 "/"


Operating clearance ratio - 0.002 "/"
Lubricant - NaK 78 at 100 F
Range of Reynolds Numbers - 1250 to 11,700 (10,000 at the 24,000 RPM design spee
Steady-state load application capability - 0 to 300 lbs. per bearing (i.e.
0 to 75 psi on projected area).

The steady-state load is applied through a hydrostatic, nitrogen fed bearing


located at the center of the rotor. Dynamic loads are applied by introducing
known unbalance weights in two planes in the rotor.

The test program calls for determination of the bearing steady-state and dynamic
load capacities, as well as of its stiffness and dynamic coefficients, from
measurements of journal center displacement and orbits over the above noted
load and speed ranges. It is the Intent of the test program to obtain this data
on the steady-state and dynamic characteristics of the bearings and rotor-
bearings system with very high accuracy, in order to establish the degree of
confidence with which the existing turbulent lubrication theory may be used to
predict alkali metal lubricated bearing performance. Specifically, any deviations
between calculated and measured performance due, for example, to fluid inertia
effects at pad entrances and exits (velocity lead losses) are to be noted, as
functions of Reynolds number and eccentricity ratio . Tilting pad bearings were
selected because their high stability makes them, prime candidates for high speed,
process-fluid lubricated turbomachinery.

The test rig is mounted in a dry box under nitrogen cover gas. The NaK is fur-
nished from a 50 GPM, 40 psi ffeK circulation assembly and gas purifier. The
assembled dry box and NaK circulation assembly are shown in Figure 9. Figure 10
is a photograph of the test rotor and Figure 11 shows one of the pillow blocks
with the bearing pads and seal rings. Figure 12 shows the assembled test rig,
but with the upper half of the cylindrical housing and the loader bearing omitted.
The hydrostatic loader bearing is shown in Figure 13. The rotor is driven through
a splined quill coupling by a variable speed (0 to 36000RPM) motor and MG set.
There are six measurement planes, one at each end of the two liquid metal lubri-
cated bearing; immediately outside the seal rings, and one at each end of the
nitrogen loader bearing. Two capacitance probes installed at 90 degree to one
another are located in each measurement plane to measure the journal displacements
and orbits.
The tests were started in October 1966 and the unit was operated first in an
"easy to use" fluid for check out purposes and later in NaK. To date, three
test sequences have been conducted in NaK. The first sequence was at speeds
up to 12,000 RPM and was made for the purpose of check out of the system, in-
cluding the NaK circulation loop, feed and drain systems, instrument performance
and others. In the second two sequences, which were conducted at speeds up to
26,400 RPM, test data was obtained first with a balanced shaft and then with an
unbalance of 0.29 oz.in., corresponding (at the highest speed of the test)to a
dynamic load of 180 lbs. per bearing (i.e. 45 psi on projected area). In both
sets of tests the static load was the 57 lb. weight of the rotor, which is
equally supported on the two test bearings. Figure 14 shows photographs of some
of the orbits at different speeds. Currently, preparations are in progress for
conducting other test runs with larger values of steady-state and dynamic loads.

At this early stage of the experimental program, it is premature to draw any


final conclusions regarding the performance of the NaK lubricated tilting pad
bearings or the degree of confidence with which their characteristics may be
predicted from theory. The early tests do, however, indicate generally good
correlation between the theoretical and test data. There has been no indication
of instability at any speed, although there is some very preliminary evidence
that at speeds higher than 15,000 RPM (Reynolds niimber = 6200), fluid inertia
effects may start to have a small, but measurable influence on the eccentricity
ratio of the particular bearings used in the tests.
3.21

REFERENCES

1. E.B. Arwas, "Topical Report on Turbulent Lubrication," Topical Report under


A.I. Subcontract N2-S9-1599 of AEC Contract AT(ll-l) GEN 8, MTI Report
64TR6 7, November 1964.

2. H.G. Elrodj Jr., C.W. Ng and C.H.T. Pan, "A Theory for Turbulent Films and
Its Application to Bearings/' Topical Report under AEC Contract AT(30-1)-3363,
AEC Report No. NYO-3363-2 5 MTI Report 65TR9, March 1965 (also published
under the same title as ASME Paper 66Lubl2, June 1966).

3. J.W. Lund et al: "Rotor Bearing Dynamics Technology," Final Reports under
USAF Contract No. AF 33(615)-1895^ U.S. Air Force Reports AFAPL-TR-65-45
Parts III and V^ MTI Report 64TR14 and 65TR15, May 1965.

4. F.K. Orcutt and E.B. Arwas, "Analysis of Turbulent Lubrication, Volume 1 -


The State and Dynamic Properties of Journal Bearings in Laminar and Turbulent
Regimes," 1st Volume of Final Report under NASA Contract NAS-w-771, MTI Report
64TR19, May 1964. (Condensed version published as ASME paper 66-LUBS-4,
(The Steady-State and Dynamic Characteristics of a Full Circular and a
Partial Arc Bearing in the Laminar and Turbulent Flow Regimes, June 1966).

5. (a) F.K. Orcutt, "The Steady-State and Dynamic Properties of the Tilting
Pad Bearing in the Laminar and Turbulent Flow Regimes," Topical Report
under NASA Contract NAS-w-1021, MTI Report 65TR32, June 1965 (condensed
version published under the same title, as ASME Paper 66-LUB-19, June 1966).

(b) F.K. Orcutt and C.W. Ng, "Steady-State and Dynamic Properties of the
Floating Ring Bearing in the Laminar and Turbulent Flow Regimes,"
Topical Report under NASA Contract NAS-w-1021, MTI Report 65TR33,
June 1965.

6. F.K. Orcutt, "Experimental Investigation of Condensing Vapor Lubricated


Thrust Bearing," TKANS. ASLE, Vol. 7, 1964.

7. P. W. Curwen, G.B. Manning, R.A, Harmon, "A Comparison of Oil and Gas
Lubrication Systems for Closed-Loop Gas-Tarbine Machinery," ASME Paper
accepted for presentation at 12th Annual Gas Turbine Conference and
Products Show, Houstin, Texas, March 1967.

8. B. Sternlicht and E.B. Arwas, 'Modern Gas-Bearings Turbomachinery - Part II,"


Mechanical Engineerings Vol, 88, No. 2, February 1966,

9. G.Y. Ono an.d 0. Deckers "Experience with Liquid-Mercury Lubricated Bearings


for Rankine Cycle Space Power Systems," Proceedings of the USAF Aerospace
Fluids and Lubricants Conference, San Antonio, Texas, April 16-19, 1963.

10. M.J. Wallace, "Summary Report of Potential Liquid Metal Bearing Materials for
SNAP 50/SFUR Pumps," Pl€AC-^68, November 1965 (prepared under Contract
AT(30-l)-2789. TID-4500, CAT.UC-25) .
R.Gs Frank (Editor), "Materials for Potassium Lubricated Journal Bearings,
NASA CR 11011, General Electric Co. Progress Report under NASA Contract
NAS 3-2534.

AiResearch Manufacturing Co. of Arizona, Report No. APS-5152-R4, Oct. 1965


(prepared under SNAP 50/SPUR Contract AF 33(615)-2289).

H.O. Slone, "SNAP-8 Development Status," Paper No. VI of NASA SP-131,


Space Power Systems Advanced Technology Conference.

F.K. Orcutt and C.H.T. Pan, "An Experimental Study of Film Rupture in
Journal Bearings with Low Kinematic Viscosity Lubricants," Topical Report
under AEC Contract AT(30-1)-3363, AEC Report No. NYO-3363-4, MTI Report
No. 65TR13, March 1965.

F.K. Orcutt and E.B. Arwas, "An Investigation of Rotor-Bearing Dynamics


with Flexible Rotors and Turbulent-Flow Journal Bearings," Part 1 -
Analysis, Design and Fabrication of the Test Apparatus. Topical Report
under AEC Contract AT(30-1)-3363, AEC Report No. NYO-3363-3, MTI Report
No. 65TR12, March 1965.
r
OIL BUFFERED SEAL-
-COMPRESSOR

TURBINE
BEAR

SEAL SAS RETURN -MAIN OIL PUMPS


TO PRIMARY LOOP

Fig. 1 Schematic Diagram of CSN = 2 Lube-Oil and Seal Gas System MTI-1604 oo
(p>-{><H (P>H>^

LOW DP
][H> DETECTOR

EXTERNAL
VI IX}-^ NITROGEN
SUPPLY

Fig. 2 Schematic Diagram of Lubrication System for the Proposed


CSG-1 Gas-Bearing Turbocompressor
r

Fig„ 3 CSN-2 Lube-Oil Distribution Panel and Lube-Oil Skid


0^

Fig. 4 Rotor and Gas Bearing Components for CSG-1 Full Scale Simulator MTI-1625

t.
Fig. 5 Bu-Mines/AEC, 24,000 RPM Gas-Bearing Turbocompressor
00

Fig. 6 Rotor, Housing and Bearing Parts for 150,000 RPM, Two-Stage
Regenerative Compressor

i.
r

. - 'i .-•;'. X •'.


•• •^~;"^f^'^'7.'. = :i; -

COMPRESSOR END BEARING COMPRESSOR END BEARING


SPEED • 18,000 R P i SPEED: 110,000 R P i
DRIVE: s.CR. POWER PACK DRIVE: i.G. SET
SQUARE WAVE OUTPUT SINE WAVE OUTPUT
I CM = 0.00025 IN. I CM = 0.00025 IN.

Fig. 7 Measured Orbits at Compressor End Bearing of Two-Stage


Regenerative Compressor

MTI-2341
o

Fig. 8 Test Rotor for HTGR Water L u b r i c a t e d Bearings

Wi
r

Fig. 9 Dry Box and NaK Circulation Loop for Liquid Metal Bearings L!
Test Program (Gas Purifier Not Shown) i^
MTI-2350
F i g . 10 Test Rotor

MTI-2344

1
«•

Fig. 11 Pillow Block, Bearing Pads and Seal Rings


MTI-2345
4>

Fig. 12 Assembled Test Rig (Upper Half of Housing and N^ Loader Bearing
Not Shown)

% .
r

F i g . 13 N„ Loader Bearing CO
U1
MTI-2347
N = 5000 RPM N = 13,000 N = 21,000 RPM
Rotor Balanced Rotor Balanced Rotor Balanced
Scale: 1cm = 0.8 mils Scale: 1cm = 0.8 mils Scale: 1cm = 0.8 mils

N = 5000 RPM N = 21,000 RPM N = 26,000 RPM


Rotor Unbalanced = 0.29 oz.ln. Rotor Unbalanced = 0.29 o z . i n . Rotor Unbalance =0.29 oz.in.
Scale: 1cm = 0.8 mils S c a l e : 1cm = 0 . 8 mils Scale: 1cm = 0.8 mils

Fig. 14 Measured Orbits


4.1

SECTION 4

TECHNOLOGY I - FUNDAMENTALS
4,

SUPER-LAMINAR FLOW IN BEARINGS AND SEALS

by

C. H. T. Pan and J. H. Vohr


Mechanical Technology Incorporated
Latham, New York

ABSTRACT

Possible fluid-dynamic processes in thin films of low kinematic viscosity fluids


are discussed and parameters governing the flow regimes are identified. Recent
analytical and experimental studies on related subjects are reviewed. Analysis
based on the "wall law" of turbulent shear flow has accounted for most experi-
mental data on super laminar thin film flows. The phenomenon of secondary
vortices is important only for relatively thick films (according to lubrication
practice). Mean inertia effects are the major unresolved items in the analysis
of bearings and seals lubricated with low kinematic viscosity fluids.

I. INTRODUCTION

Conventional fluid film bearings operate in a manner that the viscous shear
stress predominates in the fluid-dynamical process, which obeys the classical
lubrication theory originated by Reynolds (ref. 1). Reynolds' lubrication
theory is predicated on the propositions that a state of laminar flow pre-
vails and that the fluid film thickness is considerably smaller than the
characteristic dimension of the bearing surface. In recent years, trends
toward process fluid lubricated bearings bring to light the need to consider
lubrication films of fluids, which, because of a low kinematic viscosity,
cannot be adequately described by Reynolds' lubrication theory. In table I,
several typical process fluids and their kinematic viscosities are listed
for comparison with SAE No. 10 oil. Depending on the hydrodynamic and the
geometrical parameters in effect, one or more super-laminar phenomena may
dominate the operation of a process fluid lubricated bearing. The purpose
of this work is to discuss the relevance of these phenomena to lubrication
problems, to review the progresses thus far achieved, and to project future
needs of research. The fluid-dynamic point of view will be adopted, there-
for this work is as much directed to fluid seals as it is to fluid film
bearings.
CLASSIFICATION OF SUPER-LAMINAR FLOWS

A. Review of the Classical Lubrication Theory

An appreciation of the significance of a low kinematic viscosity can be


gained by reviewing the mathematical foundation of Reynolds' lubrication theory.
For the present purpose, validity of the equations of Navier-Stokes will be
presumed. Thus, for mass continuity and momentum balance of an incompressible
fluid, one finds

3U
--J- - 0 (1)
3x.

3
(a^j) (2)
^ 1 at - "j 3xj ^ij 3x
j

Here, Cartesian tensor notation xs used for brevity, curvature effects as


may be related to geometry and kinematics are understood to be implicitly
included in above equations. The applicable constitutive law is
|3U. 3U.|
'i^-^ ^ij + ^ l ^ + ii:) ^2)

It is sufficient to require that M be treated as dependent only on T and p,


and that T(x,, t) be obtainable in some consistent manner. The lubricated
surfaces are separated by a thin film of fluid, the representative film
thickness is designated as C. Let L be a typical linear dimension of the
lubricated surfaces, V be a representative velocity magnitude in the fluid
film, and AT be a representative time period. Make X2 be the spatial
coordinate in a direction normal to the lubricated surfaces, then provided

(4)

(5)

(6)
for i = 1,3

(7)
4.1.

(8)

Neglecting 0 |-r , the righthand side of eq. (2) is reduced to

^l3^ (9)
^Xj ^""i ^^"2

for i »» 1,3

In Reynolds' lubrication theory, it is assumed

so that the left hand side of eq. (2), which accounts for the Inertia forces
in the fluid film, can be neglected for i « 1,3 ; and p can be regarded as
independent of X2. In typical oil bearings

^ ^ % 100 (11)
U

f :^ io~3 (12)
LI

and

AT ^ I (13)

Thus the inequalities of eq. (10) are readily justified for an oil film.

It is seen from Table I, that the kinematic viscosity of a process fluid


is at least one order of magnitude smaller than that of oil. Thus, eq. (10)
would not necessarily be satisfied, and one can be readily convinced that pro-
cess fluid lubrication films would be distinguished by the relative dominance
of the inertia effects.

B. Flow Instabilities

The significance of the inertia effect is more than the simple prevalence
of transient and/or convective forces. Two types of breakdown can happen to a
laminar flow when the kinematic viscosity is sufficiently small.
One of these is caused by the centrifugal force of a curved layer of
flow. Its classical form exists in the annulus between an inner rotating
cylinder and an outer stationary cylinder [ref. 2]. When the rotational
speed exceeds a critical value, depending on the kinematic viscosity and
the radius ratio, a system of toroidal secondary vortices will appear as
illustrated in Fig. 1. When the radii of the inner and outer cylinders
are nearly equal, the critical speed for secondary vortices is

(14)

and the axial spacing of the vortices is approximately same as the annular
gap AR. At their first occurrence, these vortices are time independent.
Until this critical speed is reached the torque required to rotate the
inner cylinder is directly proportional to the rotational speed. Above
the critical speed, the torque will exceed this condition of being proportion-
al to the speed as shown in Fig. 2. The pertinence of the secondary vortices
in the operation of process fluid journal bearings and shaft seals is self-
evident; here the annular gap AR is the nominal radial clearance C. Secondary
vortices do not occur if the outer cylinder instead of the inner one rotates.

The second type of flow instability is due to the tendancy of a shear


flow to dissipate its kinetic energy in random fluctuations known as
turbulence. First discovered by Reynolds m high speed pipe flows [ref. 3],
turbulence is now known to exist in all types of shear flows. Without
turbulence, the laminar shear stress is linearly proportional to the mean
flow rate. With turbulence, the mean shear stress is more or less proportion-
al to the mean kinetic energy or the square of the mean flow rate. The crite-
rion for turbulence to occur is roughly

^ > 103 (15)


y
If the above condition already exists, and since the mean state of flow instead
of the random flow fluctuations is of interest in lubrication problems, eqs. (1)
and (2) should be revised to read

au.
^ . 0 ^i6>
ox.

MTI-2355
^ ir
'\ .rr ^
+ "3 i ^
.- !-^^ --'-^
„^ "i'-'d'^ <^''

where barred quantities are temporal averages and primed quantities are
the fluctuating parts. The last term in eq. (17) represents turbulent
momentum transport which causes the mean stress to be higher than the
laminar viscous stress. The criteria for neglecting the left hand side
of eq. (17) are

where C^ = , ^,7 is the coefficient of friction. C^ itself is a


^^ DVC
function of the Reynolds number ^— , and in the laminar condition eq. (18)
reduces to eq. (10).

C. Flow Regimes Map

In above discussions, super laminar phenomena in lubrication application


have been identified as

1. prevalence of inertia effects when

1 C > T 1 C ? 1 fia\

2. occurrence of secondary vortice for a rotating shaft when

and

3. presence of turbulence when

P V C , 103
y
Since C^ itself is a function of the Reynolds number, above criteria can be
pVC C C C
expressed in terms of , —, r-—, and —. A composite flow regimes map with
DVC C C C
-— as the ordinate, and either —, or 777;fr, or • as the abscissa is shown in
y L VAi K
Fig. 3. The line marking the importance of inertia effects is based on

I- £ _ 1 L_ = 0 25

C^ L C. VAT

for the Couette flow. Specifically, for -— < 2000,

C =-^
and for ^— > 2000, C is calculated according to [ref, 4], V is the velocity
f y pVC 1
y
of the moving surface.

Worthy of special notice is the fact that typical process fluid bearings
are designed with — between 10~3 and 3 x 10~3. In this range, the regime of
flow with secondary vortices but without turbulence occupies a very small por-
tion of the map. Also worthy of notice is the fact that the line marking the
importance of inertia effects is almost vertical for — — > 2000. When
inertia effects are to be considered, they include transient, convective,
centrifugal, and Coriolis accelerations.
III. FLOW WITH SECONDARY VORTICES

A. Vortex Transition in Flow with Pressure Gradients

From the flow regime chart shown in Fig. 3, one would expect that the first
transition from laminar flow to occur in bearings with C/R > 10" would be to a
form of vortex flow rather than to turbulence. It is, therefore, of considerable
practical importance to know as precisely as possible at what speeds transition
to vortex flow will occur. The vortex transition boundary shown in Fig. 3 is
for the classical case of concentric cylinders with no pressure gradients in the
flow. However, in loaded journal bearings, both axial and circumferential pressure
gradients exist. Also, in loaded journal bearings, the radial clearance varies
circumferentially. Obviously, then, the problem of the stability of flows in
journal bearings to development of Taylor vortices is very much more complex than
for the classical case of concentric cylinders.

The effect of axial pressure gradient on the stability of flow to onset of


vortices has been shown both theoretically and experimentally to be always a
stabilizing effect (Refs. 5, 6, 7 and 8). The effect is illustrated in Fig. 4
for rotating cylinders at different eccentricity ratios. The test cylinders used
in these experiments had a clearance ratio of C/Ri ~ .099 and a length to diameter
ratio of L/D = 6.12. The axial flow Reynolds number used as the abscissa in Fig.
4 is defined by
W C
(N ) = ^L-
^ Re'^ axial v

where W is a mean axial flow velocity determined by dividing the axial volume flow
rate by the annular cross-section area for the flow.

One can note in Fig. 4 that for (N„ ) . - = 0, flow stability increases with
increasing eccentricity ratio. This is due to the net effect of circumferential
pressure flows induced by rotation of the eccentric cylinder. This effect is
discussed below.

B, Vortex Transition in Eccentric Cylinders

DiPrima (Ref. 9) showed theoretically that for the case of concentric rotating
cylinders, a uniform pressure flow in the direction of rotation tends to make the
flow less stable to onset of Taylor vortices while a negative pressure flow tends
to make the flow more stable. When a cylinder rotates within a stationary, non-
concentric outer cylinder, a negative pressure flow is induced in the region of
maximum clearance while a positive pressure flow is induced in the region of
minimum clearance. To determine the net effect of these pressure flows on the
stability of the flow, DiPrima applied his abovementioned analysis to the local
flow at every circumferential point around non-concentric rotating cylinders
(Ref. 10). The well known Sommerfeld (Ref. 11) solution for journal bearings
was used to calculate the magnitude of the local pressure flows.

The results of DiPrima's analysis are shown in Fig. 5. In the region of


maximum clearance (8 = 0®), the increase in clearance with eccentricity tends to
make the flow less stable to onset of vortices whereas the negative pressure flow
that is developed there tends to stabilize the flow. The resultant effect as e
increases from zero is that the flow initially becomes less stable to onset of
vortices. However, for e > 0.67, the stabilizing effect of the pressure flow
starts to dominate and the flow becomes more stable than it was at e = 0.

In the region of minimum clearance (8 « 180°), the positive pressure flow


that is developed acts to destabilize the flow. However, the stabilizing effect
of decreasing clearance always dominates here so that flow becomes increasingly
more stable to onset of vortices with increase in eccentricity.

DiPrima's analysis of the stability of flow around non-concentric cylinders


was developed for the limiting case of C/Ri -> 0. If, in determining the stability
of flow around non-concentric cylinders, we insist that the velocity profiles be
stable at every local circumferential position, then the stability curve for 6 = 0 °
in Fig. 5 will be our minimum stability curve for the entire flow. However, for
cylinders with finite radial clearance, we know experimentally (Ref. 12) that,
when vortices develop in the flow, they wrap completely around the cylinders ex-
tending from regions where the flow is theoretically unstable through regions where
it is theoretically stable. Therefore, for vortices to develop, the centrifugal
forces acting to produce vortex circulation in any region where flow is unstable
must be strong enough to drive not only the local vortex circulation but also
the vortex circulation in the stable regions around the cylinders.
Therefore, for cylinders with finite C/Ri, we expect that at the critical speed
at which vortices will first appear in the flow would be greater than that pre-
dicted by the line for 6 = 0* in Fig. 5.
4.1.9

H In Fig. 6 are shown various experimental data for critical speed for onset
of vortices in flow between non-concentric cylinders. For these data, the point
of onset of vortices was determined by torque measurements with silicone fluids
as the test fluids (Refs. 12 and 13). As can be seen, the measured transition
speeds for onset of vortices all are higher than predicted by the minimum transi-
tion speed curve of DiPrima. Also, one can note that transition speeds become
greater as clearance ratio increases, particularly at large values of e.

The smallest clearance ratio for which experimental measurements of Taylor


vortex transition speed have been made is C/Ri ^ 0.0104. It is not established
whether DiPrima's minimim stability curve is valid in the limit as C/Rj -* 0.
However, since DiPrima's curve represents the strictest stability condition that can
be applied, one can state with reasonable assurance that Taylor vortices would
not develop in journal bearings at speeds below those given by this curve. On
the other hand, one would expect vortices to develop in bearing flows at speeds
less than those predicted by the experimental curve for C/Ri «= 0.0104.

C. Vortex Theory of Lubrication

For the typical clearance ratios found in bearings, the range of operating
conditions in which vortex flow will occur without turbulence is fairly narrow
(see Fig. 3). As we can see later, once fully developed turbulence sets in, the
transport mechanism in the flow will soon be dominated by turbulent fluctuations
and the effect of the secondary vortex flow which is present soon becomes negligi-
ble. Nonetheless, it is still of practical interest to develop a "vortex theory
of lubrication'" which can essentially "bridge the gap" between the operating range
in which laminar theory applies and the range in which turbulent theory applies.
Such a vortex theory of lubrication has been developed (Ref. 13) based on an
analysis by DiPrima (Ref. 14). In his analysis, DiPrima developed the following
theoretical relationships for flow between concentric cylinders in the vortex
regime at speeds just above Taylor critical speed.

V T (Q)
-^ = 1/2 + Q/6 + 1 --S- F(Q) G(Q) (21)
Vi

T (Q)
= 1 - Q + 1 -- ^ H(Q) (22)
pV] /h
For T > T
c
where

mean circumferential flow velocity due to sum of


m
Couette flow and pressure flow

local shear stress at surface of inner cylinder

Q 3 V /V = 3 ^ — —
•^ p^ c -^ [ 12yRi 36 Vi
V
p mean circumferential velocity due to pressure gradient
V
c mean circumferential velocity due to rotation = Vi/2
Vl
surface velocity of inner cylinder
Rl
radius of inner cylinder
R2
radius of outer cylinder
V
fluid viscosity
Vih
T local Taylor number = 4
Rl + R2
h local radial clearance between inner and outer cylinder
« C (1 + e cos 8)

local transition value of Taylor number for onset of


^c^^^
vortices. T (Q) is a function of Q, Table II

^(Q)> G(Q), H(Q) functions of Q, Table II

Relations (21) and (22) are derived for the case where both clearance and
pressure gradient do not vary circumferentially. In the case of non-concentric
rotating cylinders, both clearance and pressure gradient do vary around the
cylinders. Q, therefore, varies circumferentially. To obtain an approximate
calculation of the effect of vortex motion on shear stress and pressure gradient
in the flow between non-concentric cylinders, one can apply relations (21) and
(22) locally at each circumferential point using the appropriate local values
of Q and sum these local contributions. The procedure for doing this is describ-
ed in Ref. 13. From this procedure, one can calculate the effect of vortex motion
on the overall viscous drag associated with rotating non-concentric cylinders.

Typical results for calculated viscous drag are shown in Fig. 7 along with
some experimental measurements taken from Ref. 13. The ordinate used in this
figure is G/G , the ratio of rotational torque in the vortex regime to the torque
4.1.11

that would occur if flow remained laminar. The solid curve shown in Fig. 7 is the
theoretical curve calculated using values of critical Taylor number, T (Q), which were
determined from DiPrima's analysis for C/R^ -> 0 (see Fig, 5 ) . However, as was shown
earlier, the critical speed for first onset of vortices was found experimentally to
increase with C/Rj. To allow for this, the theoretical curve for C/R^ = 0 is shifted
horizontally to the right in Fig. 7 so as to align the point of calculated increase
in torque with the measured transition point for onset of vortices. Two such shifted
curves are shown: one for C/Ri = 0.0104 (dashed curve 1) and one for C/Rj = 0.099
(dashed curve 2). For C/R^ = 0.0104, agreement between measured values of torque
and the shifted theoretical curve are fairly good at speeds just above critical
speed. For the case of C/R-[ = 0.099, measured values of torque increase more rapid-
ly at speed above Taylor transition speed than would be predicted by the corresponding
shifted theoretical cuinre.

To determine the effect of Vortex motion on circumferential pressure profiles


around non-concentric cylinders, measurements were made with cylinders having a
clearance ratio of C/Ri = 0.0104 for Reynolds numbers up to approximately 2000. Due
to this relatively large (by bearing standards) clearance ratio, inertia forces had a
significant influence on the measured pressure profiles at the higher Reynolds numbers.
However, it was shown (Ref.13) that to first order accuracy, the effect of inertia
forces on the pressure profile would be symmetrical about the line of centers of the
cylinder while the major portion of the pressure, resulting from viscous forces, would
be antisjnnmetrical about the line of centers. Therefore, it was possible to separate
out the inertia effect from the measured pressure profile. Typical separated profiles
are shown in Fig. 8. The dashed curve (curve 2) is the measured profile including
inertia effects. Curve 1 shows the pressure profile with the symmetrical inertia effects
separated out. The inertia correction itself is plotted as the solid line through the
points designated by solid triangles. One can note that at the condition shown, the
maximum inertia correction amounts to approximately 20 % of the peak pressure.

Once the inertia effects are separated out from pressure profiles measured in
the vortex or turbulent flow regimes, one can assess, directly, the influence of
vortex motion or turbulence on the magnitude and shape of the profile. In Fig. 9
are plotted three experimental profiles measured, respectively, under conditions
of laminar flow, vortex flow, and vortex flow with perhaps the beginning of turbu-
lence. Inertia effects have been separated out from the profiles shown. The
ordinate is the dimensionless pressure PC^/6pViRi. If the flow remained laminar, all
of the profiles would follow the same curve, since this dimensionless pressure
correctly accounts for the effect of speed on the laminar profile. The fact that
the profiles measured at /i/2T* - 83 and /TJTF = 169 have larger magnitude than
that measured at /1/2T = 13.6 is a direct indication of the effect of vortex mo-
tion or turbulence. One can note that at i/l/2T = 169, superlaminar effects have
increased the peak magnitude of the experimental profile by approximately 40%.

Theoretically predicted profiles are compared with the measured profiles in


Fig. 9. In laminar flow, agreement is excellent. At /1/2T = 83 the predicted
curve based on vortex theory is seen to be greater than the measured profile.
This is due to the fact that the theoretical transition speed is less than the
measured transition. At /T/TI - 169, the measured curve is greater than the predicted
curve. In general, over the range 41 < /1/2T < 170, vortex theory seems to provide
a reasonably accurate prediction of the measured profiles.

D. Interaction between Vortices and Turbulence

The development of secondary vortex flow has a more pronounced effect on the
magnitude of pressure profiles than upon their shape. In Fig. 10 is plotted the
ratio Pmax / ( PLTmax
) V S Reynolds
•' number for non-concentric cylinders
•' at various
eccentricity ratios. Here P is the peak magnitude of the measured profile
while (P.f ) is the theoretical peak pressure that would be obtained if flow
remained laminar. The point of onset of vortices in the flow is characterized by
the ratio of Pmax to (P^) L max becoming greater than unity. Theoretical curves
based on both vortex theory and the turbulent theory of Ng and Pan (Ref. 4) are
compared with the experimental data in Fig. 10. Data shown are for C/Ri = 0.0104.
The significant thing to note in this figure is that for N ** > 2400, the turbulent
theory begins to provide a quite accurate calculation of the peak pressure, even
though it is known from visual observations that a vortex secondary flow still per-
sists in the flow. Apparently, as soon as turbulence becomes fully developed in the
flow, the turbulent transport mechanism dominates over that due to the vortex flow.

* T "^— I'T.—, r. is known as the Taylor number


V Rl + R2

Re V ^
4.1.13

The effect of vortex flow and turbulence on the shape of the pressure profiles
around non-concentric cylinders is shown in Fig. 11 where various experimental pro-
files are plotted normalized to a peak value of 1.0. As noted earlier, flow at the con-
dition /T/XI - 169 may have contained some turbulence as well as vortices, since the
Reynolds number corresponding to the flow condition was N « 1660. The turbulent
theory of Ng and Pan (Ref. 4) does predict the shape of the provile at •/TfTf = 169
more accurately than does vortex theory. However, referring back to Fig. 10 we see
that turbulent theory does not predict the magnitude of the peak pressure at this
condition (Np « 1660) as accurately as does vortex theory. Essentially, at this
condition, neither vortex theory alone nor turbulent theory alone can be expected
to adequately describe the flow.

For the case in which an outer cylinder rotates about a stationary inner
cylinder, vortex flow will not develop. In this situation, the friction factor,
A, for the flow will be a function only of the Reynolds number. A is defined as
X = 2T/PVI^ where T is the shear stress on the surface of the rotating cylinder.

In the case of rotating, inner cylinder, where vortices will develop, the
friction factor becomes a function of both Reynolds number and clearance ratio.
The dependence of A on these parameters is expressed quite accurately by the
following empirical formulas developed by Wendt* (Ref. 15),

0.073 p2_Z_S|)_l2^ (\e^~°°^ ' lO** < \ e " ^°^

A comparison of Wendt's formula with experimental data at C/Ri = 0.099 is shown


in Fig. 12. Also shown are data measured at C/Ri = 0.0104 and data measured by
Smith and Fuller (Ref. 24) for a journal bearing having C/Ri = .00293. All data
shown have been corrected to a clearance ratio of C/R-^ = 0.099 by means of Wendt's
empirical relationship which states that

(R2 - Rl) R2
X «

*The range of Reynolds numbers over which these formulae are to be used has
been modified slightly by the writer.
The dependence of friction factor on clearance ratio can be seen directly in
Fig. 13. Here are plotted curves of A vs Re for C/Ri = 0.01, 0.002, 0.001 and
0.0005, calculated from Wendt's empirical formulae. The Taylor critical speeds
for predicted onset of vortices at different clearance ratios are indicated along
the line for laminar friction factor. Also shown is the friction factor curve based
on the turbulent theory of Ng and Pan. This latter theory neglects effects of
clearance ratio since it neglects consideration of vortex flow.

It should be noted that Wendt's formulae are based on data taken with test
cylinders having clearance ratios greater than 0.095. Application of Wendt's
formulae to clearance ratios on the order of 10~3 is therefore questionable.
Hence, the discrepancy at high Reynolds numbers between the turbulent theory
curve and the curves calculated from Wendt's formulae may not necessarily in-
dicate an inadequacy of the turbulent theory.
4.1.15

IV. THIN FILM TURBULENT LUBRICATION

Analogous to Reynolds' lubrication theory for laminar thin films, turbulent


thin film lubrication theory has been formulated recently. While there were earlier
works on this topic (refs. 16, 17), the contemporary approach was pioneered by
Constantinescu (ref. 18), which invoked in substance Prandtl's "mixing length
theory." The most comprehensive treatment available to date (ref. 19) was based
on the "wall law" supplemented by Clauser's concept of "core eddy diffusivity
(ref. 20). Although there is no fundamental difference between the "mixing
length theory" and the "wall law," ref. 19 treated the buffer zone and the cross
pressure gradient effect more thoroughly, furthermore, use of "core eddy vis-
cosity" concept yields more reliable results for the large Reynolds number
Poiseuille type flows.

Neglecting inertia effects, the lubrication equation applicable to both


laminar and turbulent films in Cartesian coordinates is

S (23)
dx. 'ij dx, + 2 ^ ^ + ^ (Ph) = 0

G.. is a diagonal matrix (G.. = 0 when i 5^ j) when one of the coordinates (say,
•^ J -^ J
x,) is chosen to be parallel to the surface velocity (V„ = 0 ) . In particular,
1
G.. = TTT 5.. for the laminar film, where 6.. is the Kronecker delta. For tur-
ij 12 ij Vih h3 Sp iJ h3 Sp
bulent films, G. . depends on -77-, -TT T^—•, and —- -^ . Results computed in
' ij '^ V ' |j,V dx^' laV dx ^
ref. 19 for G . are reproduced in Figs. 14 through 17. In Figs. 14 and 15 these are
iJ Vih !h3 5p . , op
respectively plotted against —-— for various values of —- -5— with ^ = 0;
f -^ ^ °„ V I [j,V dx, dx^
in Figs. 16 and 17, — -^ is the variable with -5^ = 0. When neither
laV d x dx^
dn dp
^ nor ^ is zero, the following interpolation formula was found to be
dx-^ dx^
satisfactory:
V^h h Sp
G. . h ^P K
[iV Sj

29
G. .
/V^h
h Bp
a^ + b^, 01+
20 V (24)
'' M-V 3xj^ 3t IJ |j.V d x „

where
-1 b/a
9 tan

is the angle between the direction of the pressure gradient and the surface
velocity.
From Figs. 14 through 17, it is seen the effects of pressure gradients are
small so long as

V .2
,^, ^ 1 ^ 1 < 0.25 X 10-2
Vh jiv 3x'

In this range

G loll hl^£- hi 3sJ ^ G V4_h, 0, 0


(25)
ij \ V ' yv 3xi' )iv 3x3] ij V

which was treated in ref. 4 and the results are reproduced in Fig. 18. Using the
approximation of eq. (25) in eq. (23), one clearly sees that the equation becomes
linear in p, this is referred to as the "linearized turbulent lubrication theory."
It is quite evident that the linearized turbulent lubrication theory would be quite
accurate for nearly constant film thickness conditions. It is rather surprising that
3D
the non-linear effects of -r^— seem to remain small for a full cylindrical journal
bearing even when the eccentricity is very high as shown in Fig. 19. It appears
that the linearized turbulent lubrication theory should be adequate for all smooth-
film self-acting bearings.

The linearized turbulent lubrication theory has also been used to calculate
bearing stiffness and damping which are needed in the prediction of dynamic rotor
response. Such a prediction was corroborated by test for a rotor supported by
four-shoe tilting-pad journal bearings. Correlation between test and theory for
this bearing is shown in Fig. 20 for both static and dynamic conditions. The agree-
ment is quite gratifying.
V. INERTIA EFFECTS

All terms on the lefthand side of the Navier-Stokes equation contribute to


inertia effects. According to common mechanics terminology, they include forces
due to transient acceleration, convective acceleration, centrifugal acceleration,
and Coriolis acceleration. The latter two exist only in fluid films capable of
carrying axial loads. Undoubtedly, in the presence of inertia effects, the pres-
sure in the fluid film would not be related to shear stress only. In principle,
this situation can exist regardless whether the flow is purely laminar, laminar
with secondary vortices, or turbulent (with or without secondary vortices). An
examination of the flow regimes map quickly reveals that for a journal bearing
with a rotating inner cylinder, the inertia effects by themselves are seldom
present without either secondary vortices or turbulence. Of course, if the outer
cylinder instead of the inner one is rotating, or if the fluid film is flat, then
inertia effects can be important in a thick laminar film.

Experimental evidence of the convective inertia in the vortex flow regime was
indicated in Fig. 8 for a full cylinder. The over-all consequence is primarily a
shift of the center of pressure such that the attitude angle would be a little
larger than 90° and, to a lesser extent, the bearing force is somewhat increased.
It should be recalled that the clearance-to-radius ratio is considerably higher
than common lubrication practice.

An inertia effect is also manifested along the edge at which flow enters the
fluid film. Some data on this phenomenon was reported in ref. 16. Wall pressure
and velocity profile at the inlet edge of a cylindrical partial-arc pad were
measured for several combinations of inlet film thickness, rotational speed, and
inlet flow. This data was correlated by the formula

p = p + C ^ (26)
*^o ^a e 2

where
pU 2
p = p +
*^o "^e

p = entrance static pressure

U = mean entrance velocity


V = surface speed of rotating cylinder.

The entrance coefficient C was found to be 0.17, it is quite possible that it is


dependent on the geometry of the inlet edge. The major consequence of this inlet
effect is a shift in the apparent attitude angle of the bearing force due to the
ram pressure. This was observed in a series of tests reported in ref. 23, where
the measured bearing force correlated quite well with the linearized turbulent
lubrication theory (Fig. 21) but the discrepancy in the attitude angle is consider-
able (Fig. 22). The ram pressure at the pad inlet can be somewhat relieved by allow
ing flow bypass around the pad. This was done on a 100 deg. pad test bearing re-
ported in ref. 25 as shown in Fig. 23. The test data of this bearing agree quite

well with the linearized turbulent lubrication theory as illustrated in Fig. 24,
However, discrepancies suggesting a ram effect are still evident in both the load-
deflection and attitude angle curves, for instance, for C/R = 4 x 10-3 Re ^ 5820
and e = 0.27. Indications of increased disparity between the linearized turbulent
lubrication theory and test data for a larger C/R are also apparent in the dynamic
orbits shown in Fig. 25. It is somewhat surprising that similar discrepancy was
not apparent in the data of ref. 21. This may in part be due to a smaller clear-
ance-to-radius ratio, and also in part due to the close spacing between the con-
secutive pads.

In ref, 22, the inertia effect at an internal edge of a step pad was also
studied experimentally. It was found that a steep pressure drop occurs here. This
is probably due to the sudden acceleration of flow as it enters the step. For the
particular conditions studied, which include a fairly large gap, the land region of
the step pad practically does not support any load.

Other inertia effects (transient, centrifugal, and Coriolis accelerations)


presumably would become prominent also when the convective inertia is important.
4.

VI. CONCLUSIONS

The experience cited above can be examined in terms of the flow-regime map.
The mean
aean film thickness
thickness, C , is used to calculate the parameters which form the
coordinates of the map. C mean is defined as follows: *^
mean

Bearing Type C
mean
Full Cylindrical Mean Radial Clearance
Pocket Bearing
Land Region Mean Radial Clearance
Pocket Region Mean Radial Clearance plus Pocket Depth
Partial Arc Mean Radial Clearance
Pivoted Pad Concentric Clearance at Pivot

A flow-regime map containing all experimental information discussed above plus


several actual and contemplated design points is shown in Fig. 26, The following
remarks can be made:
1. Experiments represented by cases 6 through 9 reflect realistic
design conditions. In experiments identified by cases 10 through
12, inertia and vortex flow effects are exaggerated on account of
the large film thickness.

2. Turbulence is the dominating flow process.

3. The "Inertia border" on the map is an adequate basis for neglecting


inertia effects. Exception should be made for isolated and lightly
loaded partial arcs and sliders. For these bearings the inlet ram
effect can be quite substantial at the "inertia border."

4. For C /L < 2 X 10 3 J the thin film turbulent lubrication theory is quite


mean -^ ^
satisfactory. In fact, the linearized turbulent lubrication theory is good
enough for self-acting smooth films.

5, Inertia effects, in the presence of turbulence, are the important problems


which remain to be solved.
4,1,20

NOMENCLATURE

C representative film thickness, mean radial clearance

C entrance dynamic head coefficient

C^ coefficient of friction x/ -^ p V ^
r z m
C mean film thickness
mean

D diameter

F(Q),G(Q),H(Q) functions of Q, Table II

G rotational torque

G., admittance tensor in turbulent lubrication theory

G.^ laminar rotational torque

h local radial clearance

L representative linear dimension, pad length, length of cylinder

N revolutions per second

<«Re>axial V^^

p pressure

P pressure above ambient

p entrance static pressure

(P.,) maximum P in laminar film


L max

P max P in film
max

p entrance total pressure

Q 3V /V
^ P c
Re VC/v

Rl, R2 radii of inner and outer cylinders

i2
S Sommerfeld nximber - , L^
NyLD
t time
4.1.21
Vl h 2
t e m p e r a t u r e , l o c a l Taylor ntimber 4 —^
V Rl + R2

local transition value of Taylor number for onset of vortices, Table II

mean entrance velocity

Cartisian components of flow velocity

representative velocity, surface velocity of journal

mean circtmferential velocity due to rotation " Vi/2

mean circumferential flow velocity due to sum of Couette


and pressure flow

mean circumferential velocity due to pressure gradient

surface velocity of inner cylinder

load

mean axial flow velocity

Cartesian coordinates

temporal average of ( )

fluctuating component of ( )

arc angle

Kronecker delta

representative time period

R2 - Rl

eccentricity ratio

angle between directions of pressure gradient and tlie surface velocity

angular coordinate measured from maximum film thickness position


, 1 ,
coefficient of friction, x/ -^ P Vi^

dynamic viscosity coefficient

kinematic viscosity

fluid density

Cartesian stress tensor

attitude angle
4.1.22

•I
Subscripts

1 direction of surface sliding motion

2 normal direction

3 direction tangential to surface but perpendicular to sliding motion


4.1

REFERENCES

Reynolds, 0., "On the Theory of Lubrication and Its Application to Mr. Beauchamp
Tower's Experiments," Phil. Trans. Roy. S o c , London, vol. 177, part 1, 1886.

Taylor, G.I., "Stability of a Viscous Liquid Contained between Two Rotating


Cylinders," Phil. Trans. Roy. S o c , London, ser. A. vol. 223, pp. 289-343, 1923

Reynolds, 0., "An Experimental Investigation of the Girctamstances which Determine


Whether the Motion of Water Shall be Direct or Sinuous and of the Law of Resist-
ance in Parallel Channels," Trans. Roy. S o c , London, ser. A, vol. 174, 1883,
pp. 935 - 982.

Ng, C.W. and Pan, C.H.T., "A Linearized Turbulent Lubrication Theory," Journal
of Basic Engineering, Trans., A.S.M.E., Series D, Vol. 87, 1965, pp. 675-688.

Krueger, E.R., and DiPrima, R.C., "On the Stability of Spiral Flow between
Rotating Cylinders," Journal of Fluid Mechanics, vol. 19, part 4, pp. 528-538,
1964.

DiPrima, R.C., "The Stability of a Viscous Fluid between Rotating Cylinders with
an Axial Flow," Journal of Fluid Mechanics, Vol. 9, Part 4, pp. 621-631, 1960.

Donnelly, R.J., and Fultz, D., "Experiments on the Stability of Spiral Flow
between Rotating Cylinders," Proc Nat. Acad., Sci. 4i6, 1960

Snyder, H.A., "Experiments on the Stability of Spiral Flow at Low Axial Reynolds
Numbers, "Proceedings of the Royal Society of London, Series A, Vol. 265, 1962,
pp. 198-214.

DiPrima, R.C., "The Stability of a Viscous Flow between Rotating Cylinders with
a Pressure Gradient Acting Round the Cylinders," J. Fluid Mech. 6^, 462-468 (1959).

DiPrima, R.C., "A Note on the Stability of Flow in Loaded Journal Bearings"
ASLE Transactions, 6, pp. 249-253 (1963).

Sommerfeld, A., "Zur Hydrodynamischen Theory der Schmiermittelreibung," Z. Math.


u. Physik, Vol. 50, page 97, (1904).

Vohr, J.H., "Analysis of Turbulent Lubrication, Volume 2, An Experimental Study


of Vortex Flow and Turbulence in Concentric and Eccentric Annuli," (Vol 2 of
Final Report on NASA Contract NASw-771), May 20, 1964, NASACR-CR-54034, MTI64TR20.

Vohr, J.H., "Experimental Study of Superlaminar Flow between Non-Concentric


Rotating Cylinders," Mechanical Technology Inc. Report No. Mri-65TR43, Final
Report prepared for NASA-Lewis Research Center under contract NASw-1021.

DiPrima, R.C., "Viscous Flow between Rotating Concentric Cylinders with a


Circumferential Pressure Gradient at Speeds above Critical," ASLE Transactions
7, pp. 333-341, Oct., 1964.

Wendt, F., "Turbulente Stromungen Zwischen Zwei Rotierenden Konaxialen Zylindern,'


Ingenieur-Archiv 4, 1933, pp. 352-353.
Tao, L.N., A Theory of Lubrication in Short Journal Bearings with Turbulent
Flow," Trans. A.S.M.E., vol. 80, November, 1958.

Chou, Y.T., and Saibel, E.,M "The Effect of Turbulence on Slider Bearing Lubri-
cation," Trans. A.S.M.E., Ser. E, vol. 81, March, 1959.

Constantinescu, V.N., "On Turbulent Lubrication," Proceedings of the Institute


of Mechanical Engineers, vol. 173, 1959, pp. 881-900d.

Elrod, Jr., H.G., and Ng, C.W., "A Theory for Turbulent Fluid Films and Its
Application to Bearings," ASME Paper No. 66-Lub-12.

Clauser, F.H., "The Turbulent Boundary-Layer," Advances in Applied Mechanics,


vol. 4, pp. 1-51, 1956.

Orcutt, F.K., "The Steady-State and Dynamic Characteristics of the Tilting^


Pad Journal Bearing in Laminar and Turbulent Flow Regimes," ASME Paper No.
66-Lub-19.

Burton, R.A.,and Carper, H.J., "An Experimental Study of Annular Flows with
Applications in Turbulent Film Lubrication," ASME Paper Np. 66'-Lub-14.

Orcutt, F.K., "Investigation of a Partial Arc Pad Bearing in the Superlaminar


Flow Regime," Journal of Basic Engineering, Trans. ASME, Series D, vol. 87,
1965, pp. 145-152.

Smith, M.I., and Fuller, D.D., "Journal Bearing Operation at Super-Laminar


Speeds," Trans. ASME Vol. 78, 1956, p. 469.

Orcutt, F.K., and Arwas, E.B., "The Steady-State and Dynamic Characteristics
of a Full Circular Bearing and a Partial Arc Bearing in the Laminar and
Turbulent Flow Regimes," ASME Paper No. 66-LubS-4.
TABLE I

Kinematic Viscosity of Various Liquids

Liquid Kinematic Viscosity


* ^SAE 10
V - in^/sec
SAE 10 oil (ISO^F) 2.16 X 10-2 1.0
Lithium (406*F) 1.70 X 10~3 0.0788
Water (80''F) 1.33 X 10-3 0.0616
Silicone oil (0.65 cs) 1.01 X 10-3 0.0468
NaK (150°F) 0.99 X 10-3 0.0459
Sodium (1200'"F) 3.86 X 10-'* 0.0179
Potasium (1200°F) 3.31 X lo-'* 0.0153
Mercury (400°) 1.20 X 10-^ 0.00556
Cesium (412°F) 0.89 X 10-'* 0.00411
TABLE II

TABLE OF FUNCTIONS OF Q

Q F(Q) G(Q) H(Q) T^(Q)

-3.665 0.341 0.312 71785.6


-3.65 0.342 0.423 70784.6
-3.60 0.335 0.638 68000.0
-3.50 0.319 0.784 62867.0
-3.25 0.302 0.929 51257.0
-3.00 0.291 0.978 40943.0
-2.6 0.273 1.17 26500.0
-2.2 0.232 1.435 16250.0
-2.0 0.185 1.570 12540.0
-1.8 0.152 1.585 10000.0
-1.6 0.125 1.570 8200.0
-1.2 0.085 1.525 6150.0
-0.8 0.060 1.480 4820.0
-0.4 .039 1.45 3960.0
0 .021 1.433 3390.0
0.4 -0.0055 1.430 2930.0
0.8 -0.01 1.436 2570.0
1.2 -0.026 1.453 2300.0
1.6 -0.042 1.477 2070.0
2.0 -0.058 1.51 1880.0
2.4 -0.0735 1.548 1700.0
2.8 -0.089 1.592 1505.0
21.0 -0.873 4.926 149.0
46.0 -0.873 4.926 149.0
4.1.27

Figure 1 - Velocity Pattern in Taylor Vortex Pair


4 . 1 28

llOr

lOOi

90=

5 80'
€=0
*2
Rj = Iil2 IN.
R2 = 2.00 m

i
UJ

Figure 2 - Torque is Speed Showing Effect of Taylcr Vortices

MTI-2313
4.1.29

O
>

C/R, C/L, OR C/(VAT)

Figure 3 - Flow Regimes Map

MTI-2305
4.1.30

THEORY, €=0 (REF. 5)


® € =0
A € = 0.454
20 o €=0.707

0.
10 20 30 40 50 60 70 80
^NRi) AXIAL

Figure 4 - Critical Speed vs Axial Flow Reynolds Number

MTI-2314
4.1.31

'0 0.2 0.4 0.


ECCENTRICITY RATIO, € = e/C
lO"
BEARING CENTER

JOURNAL CENTER

ECCENTRICITY, e

Figure 5 - The Taylor Vortex Transition Value of V C jC as a Function of e for Various Values of
4.1.32
100

^ C/R, = 0.099 1

o C/Rp 0.0104 r
90 —^THEORY
i 1ii_\/i« 1
C/Ri
\jf i « i
-^ 0\j 1
(DiPRIMA, REF.IO) 1
1
1
80 1
1 s
:
V
c^70 /
1
7
/o
Qi
/ y 0

/ /
^ 0 /
o 0
>" 60 i /

/
/
/ /
» /
> /

50 /
/
9/ 4>

1^ O
y" /
y i /
/
.^ #w
Ox
.^<T
y^ /
X**" •
, > • * )^^^

40 t? N
5


/

• , ^ ^
y
30
0 0.2 04 0.6 0.8 1.0
ECCENTRICITY RATIO, €
Figure 6 - Measurements of Transition Speed for Onset of Vortices
in Flow Between Eccentric Cylinders

MTI-2304
1.8
O RUN
A KUN Z
C/R, = 0.0104
o RUN 3
1.6 €=0.51 „ V RUN 4
C/R, =0.099
6 = 0.50 ^
.X*
14- ^ - ' ^
NOTE•• CURVES 1 AND 2 ^ "
ARE THEORETICAL CURVE SHIFTED VORTEX THEORY, C / R , — 0 ^ ^ ^
^ - - ^
FOR COiPARISON WITH iEASURED "
DATA ^ ^ o
^ 1.2 1 I i . E 1 i

o
^ 7*. e^ •

0 __ m
^ ^ .—^^ ^ 4
THEORETICAL TR•ANSITION iEASURED ' iEASURED TRANSITION
r\r\ 1 HIT- r\ i r\
ruiiMi , t / r TRANSITION

0.8 ViT, =36.-1 POINT, - / I L =62.5 ^


C/R, =0.0104
-'T /£4 T" t =49.6

0.6
0 10 20 30 40 50 60 70 80 90
2C
V,C/i
y R1+R2

Figure 7 - Ratio of Superlaminar Torque to Laminar Torque,


Comparison of Experiment with Theory
MTI-2309
4.1.34

1.8
>/" "X .CURVE 1

1.4 I ^y^N\^A/CURVE 2
f

/ 1
r
®

1.0 / 1
f
> u
0.6 4/ \\

\\
^/

0. / u
V
—f-1

>
o> M
CD """"**"- ^ i ' ^
^ \ ^ /

•0.
€ = Q68,V T T = I69 , NRe=l660'll
^
JL- ^
f
/

//
•0, o PROFILE WITHOUT u y
INERTIA EFFECT ll A
^

^ PROFILE INCLUDING 11
INERTIA EFFECT /^
^

A INERTIA CORRECTION 11 \
\ \
J
{CURVE 2 - C U R V E 1) y/
\
1 1 1 1 1 1 . ^
1.8,
0 40 80 YtJO \i50 ^ ^ >2' 10 2«JO 3J?0 36 0
ANGULAR POSITION, DEGREES

Figure 8 - Effect of I n e r t i a Forces on P r e s s u r e P r o f i l e


1.6
A
/ ^^•-VORTEX THEORY,V^= 169
> 1
/ ^ORTEX THEORYVit = 83
1.2 -—^d^ V

f
//
9 ^ V - "LAilNAR THEORY (SOiiERFELD^
//
0.8 A
/.w
ir——»
kit

4 / \
/
/
.^ .(tmV
04 ^
^yf
cr. a^
^
0
to EXPERliENTAL
1 K L^
X ^y-
#^
A Y P = 69,NRe=l660
i

•^0.4 " • y ^ T = 83, NRe= 815 1


1 /C
- o
-0.8"~
LAilNAR FLOW,-/I T = 13.6
NRe = !34
1 / V }YA
\ / y/ A
€=0.68
f^ ^ / /
-1.2
F
^
'

y
/
/

-1.6
0 40 80 120 160 200 A 240 280 320 360
ANGULAR POSITION, DEGREES
4>

Figure 9 - Pressure Profiles for Laminar and Superlaminar Flow


^.1.36

30
• €=0.68
o €=051
EXPERliENTAL
2.6 A c=0.35
A e=0.20

2.2
.r
TURBULENT THEORY, €=0.68
o
^ E 1.8

TAYLOR VORTEX
S 14 THEORY, €=0.50

TAYLOR VORTEX THEORY, €=0.68


1.0
^ = 068

0.6
0.35/ >¥
CRITICAL REYNOLDS
NUiBERS
02
200 800 1600 2400 3200
REYNOLDS NUMBER, NRE = V|C/F

Figure 10 - Ratio of Superlaminar Peak Pressure to Laminar Peak


Pressure, Comparison of Theory with Experiment

MTl-2306
1.0
TURBULENT THEORY EXPERliENTAL
^
0.8 ^ L i / i T = l69 • .^1660
ND. .
VORTIXTHEORY;^ \V|TM3.6O iRg=l34
o
E 0.6 ^
Q_
^ \
Q_ 0.4
cc
• ^ l A i l N A R THEORY

k
^ \
0.2 ^F =
c/>
CO
LiJ
IE
0^
Q_
Q \ ^ ^
yj -
•0.2
—J
f = 0.68
« ^ - 0.4 \
^

o _ 0.6 \ d

/y'-
0
0.8 \
\ /%

^
H.O,
0 40 80 120 160 200 240 280 320 360
ANGULAR POSITION, DEGREES

Figure 11 - Effect of Superlaminar Flow on Shape of Pressure Profiles


k \^
^~
\ ^
V X. WENDT (REF 15)
\ >w ^ ^
0.011 /
II
^< /
\
^ 11r ^ •^Mpn
&j
/
'CORRECTED SMITH AND
^ -.
o /
jLAilNAR
^"1^^^^ - / FULLER DATA (REF. 24)
h-
^*w
X _ ^^%^ - " ^
^

Tfe» t
%
EXPERliENTAL
o C/R| =0.099
\
WkJ
p^
C ^
• C/R =00104 (CORRECTED)
!^QaOA#%
•"'-*='«*4^
0 ^ 3

0.001
100 1000 10,000
ViC
REYNOLDS' NUiBER NR^

Figure 12 - Friction Factor for Rotating Cylinders


0.01 K. ~^T^""--n^rTnn""""--^""^^^
\ \ "AD*"'
^ ^ ^ 'X^ / ^TURBULENT THE .UKl
'nV " s ^ ^ ^ NG AND PAN
.r^ -u*^
\
\ X ^•t!,
x^
1

.^ X
\ X. .•SilTH AND FULLER
\
X K C/R| = 0.00293
V -" ^ ^
" ^ w
! 1 1
r--" X,
\ .
^
' ^ - - . . ^ C/R, = 0.01
^

CM (C/R|=0.002\ ^ 'C/R| =0.002 ^"^"^^tH-L


X - -
TAYLOR\C/Rr0.00^s. fe
•-'^ 0.001 f .X f

I :RIT CAL
^ •'*-x__'''''*'*'*^'-oi_J
^, 1 ^''-^^^..^^^^^ r-^ : ^
SPEEDS C/R|=0.()005 -'-^
II
\ —
^^ 1 ^*-^
. ; : =
>
^/Ri = 0.0005-
^""

€=0

0.0001 »»«j

100 1000 10,000 100,000


Vi
NRE
Figure 13 - Effect of Clearance Ratio on Friction Factor
for Rotating Cylinders
MTI-2308
4.1.40

1x10"

^
- ^
— ^ ^

_.
^ \
1x10^ - ^

)^
[iV Ox

V
^ ^

o

0.01
X
N
TxLO^
"~-

lxl08__

0.001 1 1 _UJ_L _ _ i _ _ i _ JJ_LL _ _ J _ J _ ...XIXL


4
10 10' 10 10-
Vh
REYNOLDS NUMBER V

Fig. 14 Coefficient G , for Sp/Sx


4.1.41

0.1
1x10"^ p
4 p
1x10
^**'*'*^

^ \
llxlO^
\
X.
^
O
0.01
^
3
^p-Lix >n6
N^
P -t-. J

1x10'
^
ramiisr**®*""
X
P
, .8
1x10 —

1 L_ MM 1 1 ™UJ_L 1 1 „JJJJ.
10 10- 10 10'

REYNOLDS NUMBER = —

fig. 15 Coefficient G for Sp/Sx. = 0

MXI-2295
4.1.42

0.1 _ _ _ ^ _ _
-
-
:24
^

^^^k.
Si^

0.01
XV
^

o
— 1x10^

VX
——.^JV
IxlO''

\
———.^^
5x10^

0.001 __1_J_ 1111 1 1 _LJ_LL __J_i_ «I_LJJ_


10 10" 10 10"

REYNOLDS NUMBER = —

Fig. 16 Coefficient G for 3p/3x = 0


4.1.43

I
0.1
__
— ,
Z^Io^
LxlO^
"™"

h^ Sp 1x10^
=
3
V.
\

O
0.01
__.
1x10^
X
" ^
\
.
1 xlO'7
__ • ^

5x10'' .

.001 ____L__L_ 1 11 1 i 1 _LLLL __J_L_ _iJLLL


10 10- 10 10-
REYNOLDS NUMBER Vh
V
F i g . 17 C o e f f i c i e n t G„„ for B£__
33
^1

MTI-2299
^11' ^33

o o
I

o 11 1.11^ 11
00 / 7
^
00
\
o
1 // /^
o h- /
p
a. o
h-'
O < <
J/ /
\
o
p
<

p
/ .
V
/

^
1/V / /
y 7

t.
5x10

0 2 Z/L 80" 120 160"


200^= 240"

Axial Location Circumferential Angle (Degrees)

Fig. 19 Circumferential and Axial Pressure Distribution. Compare between Linearized


and Non-Linear Turbulent Lubrication Theory for Full Journal Bearing.

MTI-2303
46

02 04 06 08
ECCENTRCTY RJTIO igi ml ECC'IJTRCTY RATO ..11 ml

Comparison between measured a n d calcuSoted s t e a d y state Comparison between measured a n d calculated steady-state load
load capacity — m = 0 city - m = 0 5

Comparison between measured a n d calculated bearing friction


torque ~ m ~ 0

"
u ^ --.» ^ CEiTERLliE
BEARIN6
^ / e
^
3£ r ^ o
-«^ CENTERLINE
3 1 / o
SHAFT
1

1 -j—i /
'A
« 1f
s THEORETICAL CURVES AND
EXPERIMENTAL DATA POINTS
Cleardiueiatu (Cj,//0 > X 1<> ' m/ii
om 05 W 200LB
SleiideineNsiatio {I ID) - 1 (I
» m 0 W 200 LB
iVic length (8) 80deg
- - » m 0, W ZOLB
PlV( t pOMtl 111 (9p/(3) ~ OOT

0 I f 0 12
SHAFT SPEED RPM « l O '

Figure 20 - Comparison Between C a l c u l a t e d and Measured Dynamic


Load Response O r b i t Amplitude
4.1.47

lOOC ~ Linearized Turbulent 000 -• L i n e a r i z e d Turbulent


Lubrication Theory - L u b r i c a t i o n Theory
Exp. Data, Orcutt Exp. Data, O r c u t t /
- o Re = 3700 - 0 Re = i375 /
A Re = 7400 a Re = 2750 /
ORe = 11,100 D Re = 5500 « e , / r e ^ \L
«l Re = 8250 **^¥
J T^
100 100
— Re=82 i O X
/J 5500

r
OO

--
y /
J

10
3700
10
""

y J4

/
/
/

/
11,100 ^ A /
_.
— L/ /
r-
1375
/
- _ ™ ^

/
1.0 1.0
0.4 0,5 0.6 0.7 08. 0.9 1.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Eccentricity Ratio Eccentricity Ratio
C/R = 3.9 X lO"^ C/R = 2.9 X lO"-^
Fig. 21 60-Degree Partial Arc Bearing -
Load Versus Eccentricity

90^
Linearized
80*^ Turbulent
70' 0-2 Theory
— - Re = 1375
60^ Re = 14800
0.4
EXPERIMENTAL DATA
0.S5 C.8.0IL
50^ L/D = i/a p-so®
\ ">< /Ml / / — ®
I
20'
40
LB LOAD
2.9 » 10 *
0.8 ® '^°^
40^ + 20 "1 LB LOAD
t @0 C/R=
1 0 * I20J 3.9 t 10"^

30" 20° 10° 0


Fig. 22 60-deg. partial arc bearing -
attitude angle versus eccentricity
^m-

o,^^\Ki:^
N--^^ \^^

TO
C
1-1
(0

ho

O
O

H
rt
H-
OJ
I—"

S>
M
O

to

*J^ ; C ? ^ |
* / A
y
"I^ /y
^ V ^ __>w^
^^^^-^ ^2L_I
%f
THEORETICAL RESULTS > - " ••--^^ 7"^ ' — _
I3,3M
^ \ Xl665
^
a.
r —Vi
»|z . ^^ - \V ^ *<r
^ . ^ ^ « .^ ——
^^ —
^^^^ _^^^^
.4^ >-*^
/ Re C/R

/
/ j^
..ai^ X 1665 2 « 10 5
• 5820 2 « 10''
/ 0 5820 4 « lO'
/ A 13304 4 « lO'
^ f

01 02 03 04 05 06 07 0£ 09 10
ECCENTRICITY RATIO, «

M e o s y r e d o n d eoleulated sfeady-slote l o a d d a t o f o r 100 d e g arc bearing ( t / D - 1)

10^ _ — . ~-™i ^ ™„

. .^^^
-y
-- - /
— ^ - __^_,
— - -r 1 "~^"~^

-4__- .„^^„__--^
J — ^

:
\ — .

^
'? •;z=^
BIZ 1_ ^ „u ^
' ^ lOO* PARTIAL ARC
a ^ o C/R 2 «10 5 INCH/INCH
^^,„.H^"*^^!NAR FLOW 50 cs OIL EXPERIMENTAL
OATA
^
03 04 05 06 07 08 09 10
ECCENTRICirr RATIO, ,

M e a s u r e d a n d c u k u l o f e d sleody-stote l o o d data for 100 deg are


b e a r i n g ( t / D = 1)

Figure 24 - Measured and Calculated Steady-State Characteristics


of 100 Deg. Arc Bearing
4.1.50

>r Y
1 nA
.24
•i ^ 1 111 1 16
S =0.120 S=0.046 .20-
S = 0.0906 14
^ "\ s 14 *-<^
.20
-—THEORETICAL THEORETICAL .16 THEORETICAL
«^^' > — THEORETICAL •I2
--' __ X • ^

EXPERIMENTAL 16
EXPERIMENTAL EXPERIMENTAL / ./ ^^- - .-- "" r-r- ^ k EXPERIMENTAL ....--' .— ' T ^ •^.
12 s
— • .12- / / ^4'^ r^ .X" 08 v
^ A' \
j»=. M I' /' ^M
h^
.08 • " " ^ 06 06
^ ^ ' ^/
/
c
•a" .04 > \
/-r
^0 04 A> 04 ^
1
A /^ / /'
1
•1^
/ /^
1 04^
/ V _ }i f 4/
/ 02
i //
02-
// „
-2 8 -24 -2 0 -16
1
1 I-.0 8 -.04
.P^y 1.1 5 .20 .24 .28 "-.28 -24 720 -.16 i / ' / 0 8 -04 M ^ St
1
15 20 .24 .28" X-
-2
° /7- - 4 -12 -10 -08 -06 -04 -02 02 04 06 08 10 12 u
14 / • ' 18 20"
V

"-22 - 2 0 / ^4 -16 -14 -12 -10 -08 -06 -04 -02 02 04 06 08 10 12 14 //{% 2o"

--«o . - s
•^
-.08
^
/ INCHES X10''
i.^ 0 ^
-.08
INCHES X10"' // -04
<^. - y
/. / INCHES X I C ' //
/ ^
-04 i INCHES X W '

• - _ / ^ r
r
-06
-.12- -.12 ^
.M.. 1 -^
-.16 -.16- \ x? ^^ . / " -^ -+•;/ ^
-.20-
\ '^~ • - .
_- —
—o— — "f ^ \ -J .,- -•\l ^ --
\N y \ ,^
_0>i
V, ..«-^
X ' --> ~r -14

^-
Y " " INCHES X10'' ^ INCHES X 10'' T, -16 -^-
r" Y NCHES X 10"' Y NCHES X I C

Y 'r j 1 1 1 1

14
S = 0 0372
.20 -—THEORETICAL 16
S =0.0338 THEORETICAL 12 i„™ a— Re =8314
EXPERIMENTAL EXPERIMENTAL 10
—-"- - r — ^•s
— THEORETiCAL .16 .12- ^ \ C/R=2x 10"^ IN /m
EXPERIMENTAL
.12^
— -- - 0 ~ ^ V
.08- i - "' u • —
*——~.~~\-^
1
y
^' ^
1 l\
B.S'"*' f
i ^ 04
V
^ .-**" < 1 .rf7
*>'**''* <^ /»
.*^ ^ ^ -f ^ --^ 02- i /
x.^4
-.2 8 -14 -.20x^6 -.12 / -.04 k /
^'
^l^
../ /'
^''^16 .20 .24 .28
1 ) 1
- V
-20 -16 ^


/m»
,^.0 4

1
-rtft
r^"
** .12 .1B .20 .24
1 1

INCHES X10"'
yy
"-24 -22 - 2 0 . / ' ' -16 / - a

/ 1/
y ^
-10 0 8 - 0 6 -04 -02 02

"
04 06 08

^PTJ-f
y y
'O^yf^U^.' 18 20 22

INCHES X ID"'

/^ *_ ^^ /)^ INCHES X10"'


-.12
,^ \ • ^ ^^^
1 s. — -»- —--*"
S
a— «»•«• .ja- - '
\ ^' A-r
-.12- -.. , -
'\-~ .-r *- —
—— "^ -12
'r NCHES X 10"' -14
-.20
Re = 5820 Y INCHES X 10''

Y INCHES X I C '
C/R = 4x 10"^ IN./IN.

Figure 25 - Measured and Calculated Dynamic Load Orbits


for 100 Deg. Arc Bearing (L/D = 1)

MTI-2298
4.1.51

10^

POCKET

Actual and
Contemplated Bearing Experimental
10== Designs Pump Bearings
Test Refer-
Fluid Remarks ence No.
\

\ _@®LAN. 1. 3 KW HG RANKINE Hydrostatic Pads 6. Partial Arc 60° Pad Silicone Data suggests 23
Oil inlet ram effect
\ 2. SNAP 8 4 Shoe Tilting Pad 7. Partial Arc 100° Pad
' ^ ^ ' ^ ^ ^ ^ ' ' ^ Good agreement 25
with linearized
30% pad loading (a) C/R = 2 x 10"^ Silicone
turbulent lubri-
Oil
cation theory
(b) C/R = 4 X 10~3 Silicone Data suggests
3. General Atomics Oil inlet ram effect
(a) Lower Journal Hybrid Pocket
(b) Upper Journal 8. 4 Shoe Tilting Pad
(a) Unloaded Silicone Good agreement 21
4. Liquid Hydrogen Pump Oil with linearized
Hybrid Pocket
(b) 50% Pad Loading Silicone turbulent lubri-
cation theory
Oil
5. Liquid Oxygen Pump Hybrid Pocket

9. 4 Shoe Tilting Pad NaK Test in progress

10. Eccentric Cylinders Silicone Oversized clearance 8,


Oil to obtain exaggerated
Vortex flow and
inertia effects
11. Partial Arc Air Oversized clearance 22
to obtain exaggerated
inertia effects
12. Partial Arc Air Oversized clearance 22
to obtain exaggerated
inertia effects

10 ^ 10"
Figure 26 - Flow Regimes of Various Tests and Design

MTI-2296
4.

SIMULATION OF TURBULENT LUBRICANT


FILMS IN A LARGE SCALE APPARATUS

by

Ralph Burton
Southwest Research I n s t i t u t e
San A n t o n i o , T e x a s

ABSTRACT

Experiments are reported on turbulent flows in air, in a large-scale bearing


model of 6-ft dia, 2-ft length, 0.54-in.film thickness. Simulation of tilted
pads, short journal bearings, stepped pads and herringbone grooved bearings
is described. Pressure distributions are reported along with velocity and-
turbulence-intensity profiles. Wall shear stress is computed from the velocity
provile measurements. The variations of these factors are compared with available
data for plane Couette flow, wall law flow, and pressure flow in pipes, and
the relationships among these are discussed. In addition, large pressure jumps
at discontinuities are reported, and are shown to provide a major influence on
the overall pressure distributions.

INTRODUCTION

Large-scale simulation was decided upon to provide detailed information on


turbulent films, for the purpose of guiding and checking analytical techniques.
When bearing configurations are operated at turbulent Reynolds number, and have
film thicknesses of the order of 0.5 in., it is possible to measure detailed
velocity profiles and to relate these to other aspects of the flo^. One diffi-
culty with this experimental technique is that pressures produced in the flow
vary as some inverse power of film thickness, so the actual measured pressure
magnitudes are very small. Means have been devised, however, to make such
measurements (Ref. 1). Some preliminary information on both pressure distribu-
tions and velocity profiles has been reported elsewhere (Ref. 2); the present
report will be restricted principally to interesting features of the pressure
distributions J showing some of the earlier data as well as new data.
Apparatus

This apparatus was based upon a 6-ft diameter wheel, 20 in .long. About this
surface, which could be turned at speeds above 100 rpm, various stationary bear-
ing configurations could be supported, including tilted pads, stepped pads,
spiral grooves and the full bearing. End leakage could be restricted by plastic
curtains, thus making possible the simulation of long or short bearings, depend-
ing upon the presence or absence of the curtains. Reynolds number in excess
of hu /2v = R = 2500 were possible, where h is film thickness, u is journal
surface speed and V is kinematic viscosity. This is well within the turbulent
regime.

Performance of Tilted Pads

Tilted pads were 3 ft in circumferential length, 20 in. in axial dimension and


were pivoted about a film thickness of 0.5 in. at a point 0.75 ft forward of the
trailing edge. In these experiments end leakage was (a) free, (b) partially
restricted and (c) totally restricted. Pressures were measured at axial
midspan. Figure 1 shows two conditions, for a fixed angle of tilt (1.0 in.
film thickness at leading edge, 0.26 in. at trailing edge). Note that a signi-
ficant "ram effect" or "impact pressure" appears at the leading edge. In this
figure and others the dimensionless pressure P is defined as P = P/pu, , where
p is pressure, p is density and u, is half the journal surface speed u .

Figure 2 shows pressure variations for two tilt angles, and end leakage fully
restricted. Again the leading edge pressure plays a dominant role.

Leading Edge Pressures

When entrance region velocity measurements were made for different leading edge
pressures in tilted pads and stepped pads, it was found that the correlation
presented in Figure 3 prevailed. Here the dimensionless pressure is related
to the velocity in such a way that results can be correlated in terms of a
dimensionless "impact" pressure according to the equation shown. If the entering
flow had been carried at wheel speed (u , where u = 2u, ) then brought to a halt
S S DC
in a frictionless process, as on a pitot impact tube, the following relationship
would prevail:
4.

2 2 2
p/pu, = u /2u, (1)
be be

or
2
2 + - V = 2
o 2 (2)
pu, 2u,
be be

Thus the apparent dimensionless stagnation pressure of the entering flow as


described in Figure 3 is made less than if the fluid were carried at wheel
speed, is shown in Equation (2). This implies that a boundary layer exists
on the rotor surface with velocity dropping off with distance from the surface,
and with a mean velocity < u, . Even if a thin film of liquid were transported
on the journal, for the fully flooded entry condition a globule of liquid would
be expected to accumulate at the entry region, and the slowing of the liquid
(movement into an adverse pressure gradient) would be in a process of irrever-
sible mixing; and stagnation pressure would be less than the amount based on the
assumption that the flow enters with the velocity u . It is believed that the
reported data are representative of the fully flooded entry condition and are
meaningful beyond the particular experiment. It is also felt that much more is
still to be done with the entry-pressure problem.

Stepped Pads

Pressure distributions for stepped pads are shown in Figures 4 and 5. The
"unshielded" case represents unrestricted end leakage and the "shielded" one
represents zero end leakage. Of special interest is the large pressure drop
at the step where film thickness changes from 0.5 in. to 0.25 in. For the pad
with leakage the pressure drops to a sizeable negative value at this point. This
2
dimensionless pressure would correspond to -14.7 psig when (14.7)(144) = pu, (0,1)
or pu, - 21,200. For a fluid with the density of water this would occur at
be •'
approximately u = 200 ft/see. This is not an excessive speed for projected high-
s
performance applications. Thusj it would not be surprising to see cavitation under
a pad in the turbulent operating region.

Spiral Grooves
Figure 6 shows pressure tap loeatioas for a spiral groove configuration in the
large-scale appratus. Figure 7 gives pressure measurements corresponding to |
the tap locations indicated in Figure 6. The spiral angle is 45 , and the ratio
of groove-clearance to land clearance is 2:1. Note that negative dimensionless
pressures occur over the land and at one station in the groove. The maximum
negative magnitude is very near that measured for the stepped pad.

Short Bearing

Figure 8 shows a polar plot of pressure distribution around the apparatus


operated as a full bearing with unrestricted end leakage. Here the radial
clearance was 0.5 in., eccentricity was 0.5, and displacement was as shown in
the small diagram at the lower right of the figure. Most unexpected was the
fact that positive gage-pressures act in the thickest-film regions and negative
gage-pressures act on the thinner-film regions. It appears from this that film
forces act to augment eccentricity. Preliminary analyses indicate that much of
this effect is a consequence of inertial forces in the film.

Conclusion

The above examples serve to illustrate the rather sizeable interial effects in
turbulent lubricant film. These are: (1) leading edge "ram effect" which moves
center of pressure forward and otherwise significantly affects pressure distri-
bution under pads, (2) convergence effect at a step which causes a strong pres-
sure drop and may lead to cavitation, (3) negative pressure over a land of a
spiral groove configuration which can, likewise, lead to cavitation, and (4)
tendency of pressure to be high in thick-film regions and low in thin-film
regions (as in the Bernoulli effect), thus giving rise to a possible negative-
spring effect in turbulent, short, full bearings.

Preliminary analyses, which encompass the experimental thick-film case, predict


similar effects in more commonplace thin films. Though less severe in such a
case, they nevertheless appear to remain in a dominant role.

References
(1) Carper, H.J., and Burton, R.A. "A Technique for Micromanometry," AIAA
Journal, Vol. 4, No. 3, 1966.
(2) Burton, R.A., and Carper, H.J., "An Experimental Study of the Annular
Flows with Applications in Turbulent Film Lubrication," ASME Paper 66-Lub-14.
.24
LIMITED END LEAKAGE
A

P .12

Fig. 1 Pressure variation for tilted pads for unrestricted and


partially restricted end leakage
4.2.6

X
£
Fig. 2 Pressure variation for two tilt angles, for a pad without
end leakage
4.2.7

. ^ J 1 I "• "1 '' ' j

o
o
.3 o\

.2 _. _

X O
o \^
P 1 OBB \ . «•

«.j2 \y
P%c o\ 0
0 _
\D

P
-.1 —

Ho \

-.2
!

-J L 1 f

0 .1 .2_ .3 .4 .5
U2

2Ub2,
Fig. 3 Variation of pad-entry pressure with inlet velocity

MII-2111
1 1 r—
UNSHIELDED

0.540 -0.280
A R»2400
O R«f200

Fig. 4 Pressure variation under a short stepped pad


1 1 1 1 1 1 1 1 1

SHIELDED -
L.E. T.E.
.6 0.540 0.265"
- A 0.540 0.280 _
R«2400
' .4 — -

Y-"^
K
• P \ .^y^
-

I -^

I " b .2 4 .6 .8 1.0

m Fig. 5 Pressure variation under a stepped pad without end leakage

• MTI-2113
4.2.10

•I

i
t
i

6"K
Fig. 6 Layout of experimental groove and land arrangements,
showing location of pressure taps MTI-2114
SEALED

0.021 OPEN

-0.043
-0.059

-0.091
-0.080
-0.080

-0.027 1/4"
-0.043 CLEARANCE
0.289

-0.016
0.289
0.257

0.278
0.203
0.I3I

1/2"
CLEARANCE

Fig. 7 Dimensionless pressure, P, for concentric operation


4.2.12
I
\
•I
/

^ O R = 2400 O x
^

""yC^\$4^r7C\JKW
Wf ^=0^

^
7?^Xg
^
\
\

\M\yy^-^
\
\
\
\
\
>
r X x
^
::2^-^w^
x^^^^ \ J 2 ! z\^rjj\^
::3j\\^/ y^
C///
X//
1
x< 9 QO /
/ 1

^
/
> / ^
i \ '
0^

Fig. 8 Pressure distribution in a short journal bearing, for


L/D = 0.28, and eccentricity = 0.5

•I
4.3.1

Motion Picture Visualization of Laminar, Vortex and Turbulent Flows


in the Annular Gap between Concentric and Eccentric
Rotating Cylinders

by

Dr. J.H. Vohr


Mechanical Technology Incorporated
968 Albany-Shaker Road
Latham, New York 12110

ABSTRACT

A motion picture was shown of various superlaminar flow patterns occurring


in the annular fluid film between a stationary outer cylinder and a rotating
inner cylinder for both concentric and eccentric positions of the inner cylinder.
The movie contained scenes showing both Taylor vortex flow and turbulent flow.
A description of the test apparatus used in the study and descriptions of some
of the typical flow patterns seen in the movie are presented below.

DISCUSSION

The test apparatus for the visual study consisted basically of a stationary
glass outer cylinder having a 4 inch internal diameter with an aluminum, rotating
inner cylinder, 3.64 inches in diameter. The inner cylinder could be positioned
at arbitrary eccentricites within the outer cylinder by means of eccentric bush-
ings at the ends of the cylinders. A sketch of the test apparatus is shown in
Fig. 1.

The test fluids used in the experiments were silicone fluids of various
viscosities. Flow visualization was accomplished by means of fine aluminum part-
icles, suspended in the flow.

The movie presented contained scenes showing both Taylor vortex flow and
turbulent flow. Taylor vortices are a form of flow instability which arise due
to centrifugal forces. The instability takes the form of toroidal vortices spaced
periodically in the axial direction. The characteristic flow pattern in a pair
of Taylor vortices is shown in Fig. 2. In Fig. 3 is shown a photograph of non-
concentric rotating cylinders with fully developed vortex flow. Vortices are
absent from the ends of the cylinder due to the axial pressure flow there which
stabilizes the flow against development of vortices.

When the cylinders are concentric, the vortex activity in the flow is
symmetrical around the cylinders. When the cylinders are eccentric, however, the
intensity of vortex activity varies considerably around the cylinder although the
axial wavelengths of the vortices remain constant. This circumferential varia-
tion in vortex intensity is shown in Fig. 4 where a set of 18 photographs taken
at 20° intervals around the test rig are assembled into a single "unwrapped" view
of the flow around the cylinders. The eccentricity ratio for this figure is
0.475. The widest part of the annulus between the cylinders is at 6 = 0° and the
narrowest portion is at 9 = 180°. Rotation of the inner cylinder is in the dir-
ection of increasing 6. The dark vertical lines at 6 = 107°, 227° and 347° in
Fig. 4 are the tie rods connecting the upper and lower flanges of the test rig.

As can be seen, the region of strongest vortex activity is in the widest


part of the annulus and is centered at approximately 6 = 50°. Progressively down-
stream of this, as the annular clearance decreases, the vortex activity damps out
until at 9 = 180° (narrowest part of the annulus) the stream lines appear to be
all parallel to the direction of rotation. A close-up high speed motion picture
of the flow at 6 = 180° did reveal, however, that the velocity at this point
varies in magnitude periodically with Z so that there does appear to be some
vortex activity still present at this point. Downstream of 9 = 180° the annulus
begins to widen again but vorticity continues to be further damped out and reaches
an apparent minimum of strength at 9 = 230°. From 9 = 230° onward, vorticity
begins to steadily increase again as the widest part of the annulus is approached.

For concentric cylinder, higher modes of vortex wave instability occur as


speed in increased above Taylor critical speed. In these higher modes, the
vortices develop a circumferential waviness, the number of developed waves being
A

a characteristic of the particular mode (see Coles, Ref, 1). Ultimately, as


speed is further increased, turbulence sets in, and the multiple-wave pattern
of Taylor vortices begins to break down with irregular eddies. With further
increase in speed, the irregular eddying increases in intensity and appears to
become finer in scale. The number of waves in the vortex flow pattern begins to
decrease until, at a Reynolds number of 4000, the waviness has disappeared. The
flow now has fully developed turbulence but still has a basic vortex structure.
This basic vortex structure persists in the flow up to a Reynolds number of
40,000, the maximum obtained in the tests.

The sequence of events described above is illustrated by photos in Fig. 5.


Particular things to note are: the beginning of irregular eddying at N.^„ = 1080,
the lack of waviness in the vortex pattern of N_„ = 4160, the feathery appear-
ance of the turbulent vortex flow at Reynolds numbers above 4000, and the per-
sistence of a strong vortex pattern even at N_,„ = 39,800.

When the cylinders of the test rig are positioned at fairly large eccentri-
city ratios (e _> 0.5) the sequence of flow patterns that occurs as rotational
speed is increased is quxte different from that occurring when the cylinders are
concentric. When the rotational speed of the inner cylinder is increased above
the Taylor critical speed, wave instability occurs first at a position just down-
stream of the point of maximum annular clearance. Because of the fact that flow
conditions and mean flow velocity change rapidly in the circumferential direc-
tion, wave instability for non-concentric cylinders never demonstrates the clean-
ly defined sequence of modes that occurs with concentric cylinders. Instead, the
wavy instability in the wide portion of the annulus becomes steadily stronger and
more irregular as speed is increased until the flow breaks down into a pattern of
large irregular eddies which completely obscure the vortex pattern. This break-
down occurs when the Reynolds number based on the mean clearance is of the order
of 500.

The essential features of flow between non-concentric cylinders at Taylor


numbers well above critical are shown in the top sequence of pictures in Fig. 6.
The right hand picture in this sequence shows the flow near the end of the narrow
portion of the annulus. One can see the point at which the laminar flow exiting
from the narrow clearance breaks down suddenly into eddies. The middle picture
shows this transition point more directly. The last picture in the sequence
shows the eddy pattern that exists further downstream in the wide portion of the
annulus. This eddying completely obscures the vortex pattern in the wide
portion.

In the bottom three pictures in Fig. 6 is shown the flow at the same loca--
tions as in the top pictures but at a much higher Reynolds number (N = 6500).
At this condition the flow in the narrow portion of the channel appears to be
turbulent Judging from its 'grainy'" appearance. In the picture to the bottom
left in Fig. 6, showing the wide portion of the channel, one can note the re-
appearance of an axially periodic, vortex structure in the turbulent flow.

With further increase in Reynolds number,flow remains essentially the same


in appearance as it does at N = 6500. The intensity of the turbulence increas
es, however, and the vortex structure of the flow in the wide portion of the
channel becomes more pronounced.

The various flow patterns described above for both concentric and non-con-
centric cylinders are typical of those which are shown in the motion picture
taken of the flow. The motion picture is a 16mm one, approximately 500 feet in
length, with no sound track. The film was prepared by MTI for National Aero-
nautics and Space Administration under contract NASw- 1021. A copy of the film
may be obtained from

Mr. Joseph P. Joyce


Technical Manager
Nuclear Power Technology Branch
NASA Lewis Research Center
Cleveland, Ohio

A technical report on the experimental project in which the motion film was
made is listed as reference 2.
REFERENCES

1. Coles, D., "Transition in Circular Couette Flow" Navy Department Report


Division of Engineering and Applied Physics, Harvard University, April, 1964.

2. Vohr, J.H,, "Analysis of Turbulent Lubrication, Vol. 2, An Experimental


Study of Vortex Flow and Turbulence in Concentric and Eccentric Annul!
(Vol. 2 of Final Report on NASA contract NASw-771) May 20, 1964,
NASACR-CR-54034, MT1-64TR20.
4.3.6

To motor
Support
iw~m
Drive Quill
k<rtm m<^>
D
Suspending
Test Section

Eccentric Bushing
for Varying Eccentricity
of Cylinders

Outer Cylinder
2
Precision Bore
4.001 + 0.002 in I.D.
Gas Pime.
24 in. in Length Outer Cylinder
Counter Balance
Weight for Torque
Measurement
Altminum Inner Cylinder
3.641 + 0.001 in. C D .

Eccentric Bushing for


Varying Eccentricity
of Cylinders

Support -zJ~
Jluflgf

Fig. 1 Schematic of Test Apparatus


4.3.7

Fig. 2 Velocity Pattern in Taylor Vortex Pair

MTI-2104
4.3.8

•I

Fig. 3 Taylor Vortex Flow between Non-concentric cylinders, €=0.475

MTl-2105
I 4.3.9

240 ~ aO

Fig. 4 Circumferential Variation in Vortex Flow arouad


Non-concentric Cylinders, e~0.475

HTE-ZiaS
M
O

y^

a §

as

ON 03
4S>
o

o
00
O

t. -
4,3.11

•^.•i' '- -- -^J----.

•^<»^-"r ^"-^/v^

N„^ = 1570, e = 0.707

Nj^ = 6500, e = 0.707

Fig. 6 Superlaminar Flow Patterns for Non-concentric Cylinders MTI-2108


4.4,1

CONDITIONS FOR THE RUPTURE OF A LUBRICATION FILM

by

Dr. J.C. Coyne


Bell Telephone Laboratories
Murray Hill, New Jersey

ABSTRACT

A theoretical and experimental investigation of the conditions for the rupture


of lubricating films is described. The theoretical analysis deals with the
boundary conditions at film separation that are to be used with the Reynolds
Equation. A simple experiment was performed using a cylindrical reading lens
for the slider and phonograph turntable for the moving surface. The experiment
was run at very light loads and small values of the dimensionless parameter nU/T
where n is the viscosity, U is the turntable speed and T is the surface tension.
Under such conditions, the differences between the present theory for film
rupture and earlier ones were magnified. The experiment gave excellent agree-
ment with the present theory.

DISCUSSION

The location of the film-cavity interface and the pressure jump occurring across
the meniscus of this interface are of interest to lubrication engineers, since
such information bears on the proper boundary conditions for Reynolds' pressure
equation. Various boundary conditions have been proposed in the literature. The
Swift-Stieber condition which is most widely used ends the film where dp/d0 equals
zero. Birkhoff and Hays in a recent paper propose a condition of incipient counter-
flow for film separation in a diverging clearance. Other older conditions which
eliminate the negative pressures or terminate the film at minimum clearance are
still in common use.

This earlier work points out the need for determining correct boundary conditions
at the edge of a lubricating film, and has provided the motivation for the analysis
of free surface separation given in this paper. Here we focus attention on the
liquid-vapor interface of a film which is attached to a stationary surface and
* The work described in this presentation and its references was conducted by the
author at Columbia University, as part of his doctoral candidacy requirements. The
work was carried out under the guicJance of Professor H.G.Elrod,Jr, Columbia Univ.
and was sponsored by Mechanical Technology Inc.
is being swept along by an opposing moving surface. The physical situation thus ^P
described is, depicted in Fig. 1. It is one of fundamental interest, and appears
in many applications besides cavitation in bearings.

In the theory we consider a smoothly separating film, i.e. fluid velocities only
in the x-y plane. The object of the analysis was to find the shape of the liquid-
gas interface from the point of attachment out over the lower moving surface.
The fluid was assumed to be Newtonian in character with constant density, viscosityj
and surface tension. The following conditions are satisfied exactly in the analysis.
On the interface:

1. The normal stress is balanced by surface tension and cavity pressure.

2. The tangential stress vanishes

3. The interface is a streamline

4. The streamwise component of the Navier-Stokes equation is satisfied.

The only major assumption is that u can be represented by a quadratic function of


y. The constants in this quadratic representation are functions of position along
the boundary and must satisfy the no-slip condition at the moving plate and also
overall mass continuity. The details of the analysis are given in Reference 1.

Some typical solutions of film height and surface velocity are shown in Fig, 2
for different values of the parameter N. N is defined as:

N - (3UU/T)^^-^

where |a is viscosity, U is plate velocity, and T is surface tension. When (N)


is large, a single stagnation point is present at 9 - jc/2. As (N) decreases, the
ratio of asymptotic thickness h to attachment thickness c decreases, and a second
stagnation point appears at h/h = 3.0, marking the onset of flow recirculation.
As (N) further decreases, the second stagnation point moves down the surface and in
the limit as (N) approaches zero, the film becomes a meniscus closed at the lower
plate. Note that the film attains its asymptotic character at about one clearance
thickness.

For any given solution, one can recover the constants in the quadratic velocity
i ^
4.4,3

representation, and thereby construct the streamlines of the flow. This has been
done for the case of (N) ~ 0.5 in Fig. 3. The separating streamline (x = 1,0),
which divides the recirculating flow from the flow being swept away on the moving
plate, is shown. A further consequence of the assumptions is that the fluid is
stagnant along a line normal to the boundary at 0 = jt/2 . The solution therefore,
corresponds to the situation depicted in Fig. 3. When 0 is close to Jt/2, the y
coordinate samples the flow far upstream. For fully developed upstream flow, the
assumed quadratic velocity distribution was proved to be perfectly consistent with
Reynolds' equation. This consistence was one of the objects of the analysis and
was exploited in developing film edge boundary conditions for Reynolds' equation.

Some coitparisons of the theory with bubble-in-tube experiments are shown in


Fig. 4. In these experiments a narrow tube was initially filled with a viscous
liquid and thereafter partially evacuated by gaseous pressure applied at one end.
A long bubble moved along the tube expelling most of the fluid out the far end
leaving a layer of h on the interior walls (e.g. blowing beverage from a straw).
Since k was small compared to the tube radius, the differences in geometry be-
tween the cylindrical tube and the planar case of the analysis is of minor importance.
Figure 3 shows good agreement between the results of the theory and experimental
results obtained by Bretherton, Taylor and Cox over a wide range of N.

Boundary conditions for ase with Reynolds' equation can now be developed. First
the pressure jump across the meniscus due to surface tension establishes the
pressure in the film just upstream of separation. Second, the flow carried under
the meniscus must equal the upstream Reynolds' flow. This establishes the pressure
gradient at film separation. The pressure and pressure gradient boundary conditions
for use with Reynolds' equation are shown in Fig. 5, in non-dimensional form, as
a function of parameter N. The pressure gradient boundary conditions according
to Swift-Stieber and Birkhoff-Hays, which are independent of the parameter N, are
also shown. As an example of the application of the boundary condition at film
separation of Fig. 5 the pressure profile for an infinitely-wide, parabolically
varying clearance was calculated and then plotted in Fig. 6. These pressure pro-
files are for the case ofa fully flooded inlet, that is, the film is assumed to
begin infinitely far upstream. The abscissa 0 of Fig. 6 is the x-coordinate,
distorted such that 0 = 90 corresponds to x - oo and 0 = 0 corresponds to the
point of minimum film thickness. Note the relatively large sub-ambient pressure
region associated with the pressure profiles for small (N) . Wl^

The object of the experimental portion of the investigation was to test the
validity of the boundary conditions derived from the theory. The experiment
was of simple design requiring only measurements which could be made easily
and accurately. The basic set-up consisted of a cylindrical slider bearing
riding in a fluid film. Measurements were made of the load on the slider and of
the location of the inlet and exit film boundaries. For comparison, Reynolds'
equation was solved for the geometry of the slider using the boundary conditions
of the theory as well as those of Swift-Stieber and Birkhoff-Hays. The re-
lationship between the load and the location of film separation was computed for
nearly flooded inlet conditions. The experiments were run for very light loads
and anall values of (N). Under these conditions, surface tension effects become
important and the present theory shows the greates t disagreement with others in the
literature. However, it should be emphasized that under many conditions of load
and N, the various results do not differ substantially.

Figure 7 shows a schematic sketch of the experimental apparatus. The slider was
a one inch length cut for a cylindrical glass lens of two inch radius. This
section was then cut in half. The moving surface was a glass disc placed over a
phonograph turntable. The lubricant was Dow Corning silicone oil. The slider was
attached to the end of a phonograph tone arm by means of a dial indicator attachment.
The tone arm had a point pivot which allowed free motion of the arm in all planes.
The turntable speed was first set to give a desired value of N. The ball and socket
joint of the dial indicator attachment was used to set the slider at any arbitrary
value of the angle J. The angle 7 was measured optically making use of a mirror
attached to the upper surface of the slider. It should be noted that the angle 7
which is measured optically bears a geometric relationship to the angular distance
between the point of minimum film thickness and the edge of the glass slider.
Therefore, when film separation coincides with the edge of the glass slider, the
measurement of 7 is tantamount to measuring the location of film separation. The
sliding weight on the tone arm was used to load the slider to the point where the
film was just on the verge of receding from the slider edge. Fine adjustment of
load was made by lowering a very light chain onto the slider surface. The location
of film inlet was measured by means of small scale attached to the side of the fQI
slider. This film inlet location measurement does not require too much accuracy
4.4.5

since small deviations from the fully flooded inlet condition result In small
differences in the load carrying capacity of the slider.

As load was applied to the slider, a point was reached where the film began to
recede upstream from the slider edge. This recession was always accompanied by
some flow in the lateral directions giving the familiar periodic striated flow
pattern observed by other investigators. The edge of the lens-slider therefore
served two purposes: one to stabilize the separating film and therefore force
compliance with the planar-flow of the thereotical model; and two to provide
a means of locating the film edge. Figure 8 shows the recession of the film
from the slider edge and the start of the striations for the case of N = ,5.
With 1.6 grams on the slider the film is separating from the slider edge. With
an additional .2 grams, the film begins to recede upstream and the striated flow
pattern is initiated. With 3.6 grams on the slider, the striated flow pattern
is fully developed.

The experimental data for N = 0.5 are plotted in Fig. 9 along with the predictions
of the present theory, the Swift-Stieber and the Birkhoff-Hays theories. Along
side of each data point is recorded the observed location of film inlet 0,. (9 equals
o
90 corresponds to a fully flooded inlet). In all theoretical predictions, the
pressure at inlet is assumed to be zero, and no correction has been made for the
weight of the fluid. The data are in best agreement with the present theory. However,
agreement becomes gradually poorer when the load on the slider is reduced. For
such lightly loaded conditions, the theory underestimates the load carrying
capacity of the slider. In Fig. 10 the same experimental data are compared with
the theory where the inlet pressure has been corrected for the hydrostatic head
Pgh.I a small correction has also been applied to account for the weight of the
fluid. The agreement between theory and experiment is excellent. Experimental
results for the case (N) =1,0 also gave excellent agreement with the theory.
For more detailed description of the experiment, discussion of experimental errors
and the calculation of load carrying capacity based on the boundary conditions of
the present theory, Ref. 2,

In summary, the main findings of this investigation are as follows:

1. In the absence of inertia or gravity effects, the shape of the liquid-vapor


interface of the separating film depends solely on the paramter N.

2. For (N) smaller than approximately 1,34, a stagnation point appears on the
liquid-vapor interface at a height of three asymptotic film thicknesses,
marking the onset of flow recirculation just upstream of the separation point.

3. The separating film attains its asymptotic character in approximately one


clearance thickness; the flow upstream of separation approaches the conditions
of Reynolds' flow also in the order of single clearance.

4. The proper boundary conditions is to be used with Reynolds' equation


depend on the parameter N.

5. The pressure gradient in the direction of flow just upstream of separation


must be positive; a subambient pressure region must exist just upstream of
separation.

References:

1. J.C Coyne acd H.C.Elrod J. , "Ccndirious for the Rupture of a Lubricating


Film," MTl Report 65TR58, December 1965.

2. J.C„ Coyne and H.G. Elrod, Jr., "Slider-Bearing Test of a Theory For Film
Rupture or Separation" MTI Report 66TR345 June, 1966.
^^>/^^^ R^TE

Fig. 1 Physical Situation


Fig. 2 Typical Curves of Film Height to Attachment Height Ratio (h/c)
and Fluid Surface-Velocity to Plate-Velocity Ratio (v/U) for
N = -2.0, -1.0, -0.5, and -0.2; in the Absence of Inertia and
Gravitation Forces. Stagnation Points (S.P.) are Shows at c/h = 3.0
00
MOVING- SURFACE

Fig. 3 Theoretical Streamlines and Fluid Velocity-Profile, Upstream of the


Separating Film, for N = -0.5, in the Absence of Inertia and Gravi-
tation Forces. a:X = 1.05; b:X = 1,08; c:X = 1.18; d:X 1.31;
e:X = 1.00; f:X = 0,95
42 -

\
27 - XBRETHERTON'S
\ DATA
A PRESENT THEORY
15- 0 MODIFIED TAPE THEORY

9 - V
6 - TAYLOR'S\
DATA \ .

^^^"-\ COX'S
3 - ^^5^___ DATA
« ^ A
2 - o

1 1 I I I 1
0.2 •0.5 -1.0 -1.5 -2.0 -5.0

Fig. 4 Comparison of Bubble-Experiment Data with the Present Theory and also
with the Tape Theory which is Modified by Retaining the Demoninator
in the Curvature Term
.001 190
R='ii

APPROACHES-1
AS N-*0

R=10
5

APPROACHES -^TIS
AS N — 0 "^^

.5 1,0 1.5 2.0 5.0


N

Fig. 5 Non-Dimensional Pressure and Pressure Gradient to be Used as


Boundary Conditions with Reynold's Equation. ^^ ^ ,3M.UV 1/3 pUh
N R =
crFHAYS(N'=-1.34)"

N =-0.2

-60" -30 30 60
Fig. 6 Pressure Profiles for an Infinitely-Wide, Parabolically-Varying
Clearance (h = h + T^IIT) where 2r/h = 900. The Pressure Profiles
were Calculated using Boundary-Conditions in Accordance vith the
Present Theory.
r

BALANCE ABOUT
X AXIS

4:^
Fig. 7 Sketch of the Apparatus

U3

MTI-2166
4.4.14

•I

(a) 1.6 grains

(b) 1.6 + 0,2 grams

(c) 1.6 4- 2.0 grains

Fig. 8 Start of the Film Striations for Medium Load, N =» -0.5


The force of the unloaded slider on the turntable was
1.6 grams. Loads were increased: (a) 1.6 grams (b) 1.8 grams
(c) 3.6 grams
MTI-2167
Fig. 9 Gamma (7) versus Load Data for N = -0.5, |J.U = -0.76 dynes/cm, compared
with the uncorrected theory (solid lines: 1 = -89 deg., -85 deg. and
-80 deg.) and other theories (dashed lines). The inlet film location
0, is indicated alongside the corresponding data point. f"
I-"
Ui
MTI-2168
4>
4>

Fig. 10 Gamma (7) versus Load Data for N = -0.5, i-iU = -0.76 dynes/cm, compared
with the corrected theory (solid Imes: ¥, = -89, -88, -87, -85, and
-80 Deg,) The first digit nSii of 9. alongside the data has been omitted.

t.
5.1

SECTION 5

TECHNOLOGY II - SEALS
5.1.1

EXPERIMENTAL AND THEORETICAL STUDY OF THE VISCOSEAL


by John Zuk, L. P . Ludwig, and R. L. Johnson
Lewis Research Center
National Aeronautics and Space Administration
Cleveland, Ohio
SUMMARY
Experimental studies were conducted on pressure patterns, gas inges-
tion, and sealing capacity of viscoseals. Oil, water, and liquid sodium
were used as the sealed fluids. Results indicated that when the groove-
land pairs connected directly with the pressurized cavity (without an
intermediate circumferential groove), an end effect, or ineffective seal
length, was evidenced by a sharp decay of pressure along the land leading
edge. Results also disclosed that a helically grooved rotor viscoseal has
increasing gas ingestion rates with increasing Reynolds number when the
viscoseal liquid interface becomes unstable. However, if the helical
grooves a r e in the housing (smooth rotors), gas ingestion could be elimi-
nated by increasing Reynolds number Secondary grooves on viscoseal
land areas improved sealing capacity and reduced power absorption. So-
dium was sealed at 300° to 625*^ F with negligible liquid loss.
From a theoretical analysis, a set of two-dimensional equations (that
includes both convective inertia and viscous forces) were formulated for
a quasi-two-dimensional flow field in parallel groove geometries which
can be used as a mathematical model for the viscoseaL The mathematical
model consisted of an infinite smooth flat plate moving relative to a fixed
infinite parallel groove-ridge surface with the clearance filled with an in-
compressible, homogeneous, Newtonian fluid. A modified Reynolds num-
b e r evolved that is based on a groove-ridge characteristic length parallel to
the smooth plate velocity vector. Application of the modified Reynolds
number to published experimental results reveals that a significant im-
provement (previously attributed to onset of turbulence) in bearing num-
ber or sealing coefficient occurred when the modified Reynolds number
was near unity; therefore, it is argued that the strong influence of con-
vective inertia, and not turbulence, is the main cause of this improve-
ment. For this reason a turbulent flow model should include the convec-
tive inertia effects. The consideration of convective inertia effects r e -
veals that fluid density p, aspect ratio h/h , and number of groove-
ridge pairs N a r e additional parameters of p r e s s u r e generation depen-
dence that a r e not revealed by creeping flow analyses. The analysis
presented herein provides a unified approach which was previously
covered by the creeping flow and the semiempirical turbulent flow anal-
yses. No restriction was placed on the net flow; therefore, the analysis
applies equally to hydrodynamic seals, bearings, and pumps. An arbi-
t r a r y groove-ridge shape is also permissible in solving equations. Con-
sideration of convective inertia effects suggests the principle of viscoseal
axial p r e s s u r e generation and attainment of zero leakage.

INTRODUCTION
The viscoseal is a special case of a general class of parallel groove
geometries which have many potential applications for use as p r e s s u r e
generation devices. For example, the parallel groove has been applied
in spiral groove thrust bearings (ref. 1), spherical and conical spiral
groove bearings (ref, 1), helical groove pumps (ref. 2), as well as heli-
cal groove seals (viscoseaL ref. 3).
The viscoseal is potentially useful for sealing liquid metals, such as
mercury, potassium, and sodium, in space electric-power generation
systems that require 1 to 3 years of unattended operation and near zero
leakage (refs, 4 and 5). As pointed out in references 5 and 6, the visco-
seal has inherent reliability and long life because of the absence of solid
surfaces in rubbing contact.
To date, analyses of parallel groove geometry have been limited to
creeping flow solutions where convective inertia effects have been ne-
glected, Convective inertia is the fluid mass acceleration due to a spa-
tial velocity change, for example, a change in velocity due to an obstacle
in the flow path. This is contrasted to an inertia effect caused by a tem-
poral velocity change, for example, a variation in velocity at a point in a
bearing due to shaft runout (which is a function of time). The creeping
flow analysis is restricted to flows where the modified Reynolds number
(see DISCUSSION) is much less than 1; that is,

Re = R e ^ l^\ « 1

or
Re"< 0.01
where

Re, ^ U L ^ p U L

£ < i
L
Creeping flow analyses for pumps, screw extruders, seals, and
thrust and journal bearings have been published by many authors ( e . g . ,
refs. 1 to 26). Appendix C contains an outline and discussion of the pub-
lished creeping flow solutions. The main criticism of the creeping flow
solution (generally used in the laminar flow regime) is that its validity is
restricted to Reynolds numbers below most engineering applications.
A number of semiempirical and turbulent flow formulations have been
devised (refs. 4, 18, 27, and 28) for operation in a regime designated as
turbulent, but they neglect the convective inertia t e r m s which are impor-
tant physically. The main criticism of these semiempirical analyses and
turbulent flow model is that the existence of the turbulent regime is a s -
sumed. Turbulence, however, may not be necessary to explain reported
experimental results and may not even exist. Even if turbulence does
exist, convective inertia effects a r e thought to be far more significant.
Kettleborough (ref. 29) numerically analyzed the slider bearing with in-
ertia, turbulent, and viscous t e r m s considered. When inertia only was
considered, the results were in qualitative agreement with published
(turbulent-attributed) slider bearing experimental results. Kettleborough
concluded that the turbulence term did not appear to greatly affect the
operation of the slider bearing,

Golubiev (ref. 24) applied centrifugal pump similarity laws to the


helical groove seal (including operation where the rotor and housing a r e
both grooved). This case considers an inviscid fluid (viscous forces ne-
glected) whereby the pressure generation is found from Bernoulli's equa-
tion modified by a geometric constant, where AP varies a s the square of
the velocity. The viscous force cannot be neglected for two reasons: (1)
the viscous force causes fluid motion, and (2) the viscous force contributes to
maintaining the generated pressure gradient. Unfortunately, the creeping
flow optimum geometry has been extended to be valid for operation beyond
the creeping flow regime. Since the convective change is the main
driving force, a complete reevaluation of the optimum geometry including
the groove-ridge shape must be investigated before conclusions on visco-
seal performance can be made. Experiments show that the optimum
geometry changes for flow beyond the creeping flow regime (ref. 27).
Several authors (refs. 20 and 21) reported that the viscoseal had a
certain ineffective length at the high-pressure end of the seal. This has
been called the end effect, and it has been suggested (ref. 20) that this
ineffective rotor length be substracted from the calculated wetted-seal
length in order to compare theory and experiment. Muijderman (ref. 1)
derived theoretical p r e s s u r e patterns for the spiral groove geometry on
plane surfaces and developed an analytical method of correcting for the
end effect. However, p r e s s u r e patterns in the viscoseal and end-effect
alterations of these patterns have not been determined experimentally.
Existing experimental data consists of average pressure measurements
from which the existence of an end effect is deduced.
Gas ingestion from the low-pressure to the high-pressure end of the
seal is reported to have occurred when sealing potassium (ref. 4) and
when sealing water (refs. 5 and 27). Although this gas ingestion charac-
teristic may present no problem with a vacuum on the low-pressure end,
it is highly undesirable in other potential applications because of con-
tamination and/or detrimental effect on sealing capacity. Reference 24
reports that in attempting to seal a liquid, the viscoseals (combining
helical grooves in rotor and housing) worked on an air-liquid emulsion,
and reference 25 reports that viscoseals (helical grooves on shaft)
operate with a mixture of oil and gas. These modes of operation could be
due to gas ingestion. Leakage observations reported in reference 21 on
"seal breakdown" and in reference 4 on "secondary leakage" could result
from gas ingestion into a closed cavity.
The objectives of these studies were:
(1) Study experimentally the p r e s s u r e patterns in the viscoseal with
particular emphasis on the pattern near the seal end (end effect)
(2) Investigate experimentally gas ingestion and sealing capacity of
the viscoseal
(3) Develop a physical flow model for a theoretical analysis of the
viscoseal
The data and analysis contained herein is covered in detail by the
authors' papers (refs, 5, 6, and 30).
SYMBOLS
a ridge or land width
a' ridge or land width parallel to viscoseal centerline
a acceleration vector
b groove width
b' groove width parallel to viscoseal centerline
C specific heat of fluid
C specific heat of fluid at constant p r e s s u r e
c clearance between ridge and smooth flat plate, or radial clearance
e unit normal directed toward axis
e. unit pormal tangent to trajectory
F body force in x-direction
2%.

F body force in y-direction


F body force in z-direction
G dimensionless factor, function of helical geometry
g acceleration of gravity
h step or edge height (or groove depth)
k thermal conductivity of fluid
L characteristic length, length of groove-ridge pair in smooth plate
relative velocity vector direction
L' axial length, normal to plate velocity
N number of helix starts
n integer
AP p r e s s u r e differential
P* static p r e s s u r e at reference state
P static cavity p r e s s u r e
Pr Prandtl number
Q net volume flow rate
Q* heat source energy
O volume flow rate normal to plate relative velocity direction
Qt volume flow rate in plate relative velocity direction
q fluid velocity in x-z plane
R radius of rotor
Re Reynolds number
Re* modified or reduced Reynolds number
T* temperature at reference state
t time
U smooth flat-plate velocity or rotor surface speed
u velocity in x-direction
5.1.8

V characteristic velocity in the y-direction


V fluid velocity vector
V p r e s s u r e flow velocity in x-direction
p, X
V _ p r e s s u r e flow velocity in z-direction
p, z
V velocity in y-direction
w velocity in z-direction
x coordinate along ridge-groove
y coordinate across film (between plates)
z coordinate a c r o s s ridge-groove
a angle between relative velocity direction and parallel groove-
ridges or helix angle
C film height
T] coordinate normal to plate relative velocity vector
T - Tg
9 temperature ratio = —-
TM"TS

A bearing number
AG empirical sealing parameter (defined in ref. 5)
II absolute viscosity of fluid
p kinematic viscosity of fluid
p* kinematic viscosity of fluid at reference state
I coordinate in direction of plate relative velocity vector
p density
p* density at reference state
ff shape parameter
# dissipation function
(p angle traversed in polar plane
V vector differential operator Del or Nabla
2
V Laplacian operator
Subscripts:
c based on clearance or film thickness
g groove
L based on characteristic length
M moving surface
n normal direction
r ridge or land
S stationary surface
t tangential direction
z across ridge-groove
e plate relative velocity direction
f] normal to plate velocity direction (axial length)
APPARATUS AND PROCEDURE
Experimental Study of P r e s s u r e Patterns
Figure 1 shows a schematic drawing of the experimental apparatus
and viscoseal assembly used to study end effects and pressure patterns.
The rotor of the viscoseal is attached to the power input shaft which is
supported by externally pressurized gas journal bearings. Axial thrust
due to sealed fluid cavity p r e s s u r e against the rotor is resisted by the
externally pressurized gas thrust bearing. The power input shaft is
driven by a variable-speed electric drive and step-up transmission. A
magnetic pickup monitors the shaft speed. The viscoseal housing is at-
tached to a support shaft and roller bearing assembly, which permits
axial adjustment. A pump pressurizes the seal housing cavity and circu-
lates a cooling fluid flow from the reservoir, through a heat exchanger, to
the test cavity and then back to the reservoir. This cooling flow is neces-
sary to control fluid film temperature in the viscoseal. The steel rotors
have a centerline average surface finish of 20x10" inch and a r e mounted
within 0. 0002 inch of the total indicator reading. The viscoseal housing
is constructed from acrylic plastic and contains thermocouples and p r e s -
sure taps as shown in figure 2. P r e s s u r e taps, in a line parallel to the
axis, a r e located at groove edges and midland and midgroove positions;
pressure taps a r e also located near both edges along one groove. Two
sets of four pressure taps (90^ apart) were used to aline the housing with
respect to the rotor within 0. 0004 inch as determined by calibration.
Thermocouples were located flush with the bore and groove root in a line
parallel to the axis and at each midgroove and midland position. Figure 2
also gives the groove dimensions used in this evaluation. The housing had
1° helix angle. Groove widths of 0. 19 inch and
a 2. 001-inch bore and a 14-^
land widths of 0. 13 inch were provided by using five helix grooves (five
starts). These geometric proportions, which are based on the optimum
relations given by reference 19, included a 0. 015 inch groove depth and a
rotor radial clearance of 0.005 inch.
The kinematic viscosity-temperature relations for the mineral oil used
were found to be the same before and after use in the test. These values
can be found in table I.
Experimental Study of Gas Ingestion and Sealing Capacity
In studies on sealing water (fig. 3), the water p r e s s u r e and flow to the
pressurized cavity (at the viscoseal high-pressure end) were controlled by
throttling valves on the inlet and outlet water lines. This arrangement
provided temperature control of the sealed fluid. A transparent housing
permitted visual observations on the rate of gas flow out of the pressurized
cavity; and, in studies on grooved rotors, a stroboscope permitted obser-
vations of the gas ingestion process in the rotating grooves.
Figure 4 is a diagrammatic sketch of the experimental apparatus used
for evaluating viscoseals in sodium. The temperature was maintained by
a furnace (resistance heaters) surrounding the viscoseal housing. The
temperature was monitored by thermocouples attached to the seal housing
outer diameter and by one thermocouple submerged in the sodium in the
pressurized cavity.
The enclosure surrounding the viscoseal assembly is pressurized
with argon (after prior evacuation) to slightly above ambient to insure
exclusion of air. P r i o r to the introduction of liquid sodium, the input
shaft is set at some fixed speed. Sodium is introduced by pressurizing
the reservoir and venting the leak detector tank (see sodium supply system
in fig. 5). The sodium first fills the pressurized cavity (viscoseal high-
pressure end) and then is allowed to fill the leak detector tank to a p r e -
determined level. The leak detector tank float displacement is monitored
by recording the differential transformer output that is produced by the
transformer core attached to the top of the float. Thermocouples placed
at different levels in the tank provide calibration during the filling process.

By holding the reservoir temperature at 220° F, a low degree of oxide


solubility was maintained. The 20-micron filter (stainless steel) inside the
tank allowed 220^ F filtration. A second filter external to the reservoir
(5 microns) provides additional filtering at approximately 300*^ F. Sodium
is partially removed from the system by pressurizing the leak detector and
forcing the sodium back into the reservoir. By alternately filling the
system and then partially returning the sodium to the reservoir, the s o -
dium can be recirculated and refiltered. This recirculation provides a
hot flush or cleaning action on the viscoseal assembly before a run. (All
transfer lines and the test section a r e held at 300° to 500° F during this
recirculation process.) For all sodium studies the leak detector tank was
held at 300° F, and any seal leakage resulted in a drop of liquid level and
float position. Gas ingestion into the pressurized cavity was indicated by
a rise in sodium level in the leak detector.
Sealing capacity, when sealing sodium, was obtained by increasing
cavity pressure until the leak detector indicated leakage. Line thermo-
couple readings were also found to be sensitive leak indicators. Visco-
seal geometries employed in this evaluation a r e shown in table II. The
geometry near the optimum (derived in ref. 19) was taken as the basis
for comparison, and is listed in the first row of table II as geometry 1.
Secondary grooves were added to the lands of this basic geometry to form
the second geometry evaluated. For the third geometry, a ratio of groove
to land width of 1. 90 was selected; and for the fourth geometry, secondary
grooves were added to geometry 3, Geometry 5 is the internally grooved
housing having the same groove dimensions as geometry 1. The rotor
outer diameters were 1. 993 inches, and the housing bores were 2. 001
inches.

EXPERIMENTAL RESULTS AND DISCUSSION


P r e s s u r e Patterns
Figure 6 shows an experimentally determined pressure pattern in a
viscoseal composed of an internal helically grooved housing and a smooth
5.1.13

h
rotor arrangement. In a plane orthogonal to the rotor centerline, the
p r e s s u r e increases a c r o s s the groove (from point A to B) in the direction
of rotor rotation and decreases across the land (from point B to C). The
increase and decrease in p r e s s u r e repeat for each groove-land pair and
produce a saw-tooth p r e s s u r e profile in the orthogonal plane. The p r e s -
sure increases from the low-pressure end (plane 3) to the high-pressure
end (plane 1), and the p r e s s u r e pattern has a helical twist corresponding
to the helical grooves. (For a rotating grooved shaft this pressure pattern
would be rotating.)
The pressure gradients in the planes orthogonal to the rotor center-
line a r e essentially linear for axial positions not influenced by the end
effects. Figure 7 shows typical experimental results. These pressure
profiles a r e the same a s those for the orthogonal planes shown in figure 6
except that a single groove-land pair is unwrapped to form a plane figure.
(Neglecting curvature is not a significant e r r o r since the ratio of rotor
radius to groove depth is 65 to 1.)
In these orthogonal planes, the ratio of pressure to cavity pressure
P/P increases linearly over the groove and then decreases linearly over
the land. This linear relation was found to hold to the maximum test speed
of 6000 rpm, which produced a modified Reynolds number (Re*) of 0. 25.
End Effect
"" When the helical grooves connect directly with the pressurized
cavity, the p r e s s u r e developed along the land leading edge decays sharply
near the high-pressure end of the seal. Typical results a r e shown in fig-
^^ ure 8 for 1000 and 6000 rpm. The pressure pattern at the end was essen-
tially the same for the speed range investigated (1000 to 6000 rpm) except
that the results for 5000 rpm (not shown) and 6000 rpm start to show a
slight shortening of the end-effect length. This shortening is probably due
to convective inertia effects. In all cases, the end-effect axial length was
approximately equal to one-half the axial groove width. This suggests that
end-effect length could be taken to one-half groove width without significant
error. The pressure decay within the groove is due to the equalization of
the land leading- and trailing-edge pressure, since a common p r e s s u r e
must exist in the cavity.
Gas Ingestion and Sealing Capacity Observations
When Sealing Water
Grooved housing. - The series of photographs in figure 9 shows the
gas ingestion process when sealing water with a grooved housing and
smooth outer diameter rotor. At a modified Reynolds number (Re*) of
1. 75 (2000 rpm), no gas ingestion is evident, the water film is clear, the
interface is stable, and the scavenging length is nonwetted.
At Re* = 2. 58 (3000 rpm) the film had large gas pockets, gas was
ingested as evidenced by air bubbles passing out of the transparent vent
line, and the sealing capacity (parameter, AG) was lower than that at
Re* = 1. 75 (2000 rpm). Various degrees of gas ingestion and film rupture
a r e evident for Re* -^ 3. 50 (4000 rpm) to 6. 10 (7000 rpm). At Re* = 7. 00
(8000 rpm), no detectable gas ingestion rate was evident, and gas ingestion
was not detectable from Re* =• 7. 00 (8000 rpm) to the highest speed evalu-
ated, which gave Re* ^ 10. 84 (12 000 rpm). At Re* = 7.00 and above,
the seal wetted length had two distinct regions (fig. 10(f)): a gas-liquid
region, and a region of homogeneous liquid at the high-pressure end.
5.1.15

h
Between Re* = 3 . 5 and 6. 10, it was observed that the bubble size
progressively decreased when moving from the low-pressure end to the
high-pressure end. It is probably that with longer seal lengths the c o r r e -
sponding higher p r e s s u r e s would reduce the bubble size sufficiently to al-
low escape over the lands, and thus gas ingestion would be eliminated. It
was observed that concentricity of rotor and housing markedly affect gas
ingestion occurrence.
Grooved rotor. - In a series of visual observations using a grooved
rotor, the gas ingestion started at Re* = 1, 9 and increased in rate with
increasing Reynolds number (maximum investigated, Re* = 10.84). Below
Re* = 1 . 9 the liquid-to-gas inteiiace was stable and no gas ingestion was
observed.
Gas ingestion mechanism. - These observed gas ingestion character-
istics of the grooved rotor and the grooved housing are illustrated in fig-
ure 10. The probable mechanism producing the difference between the gas
ingestion process of the grooved rotor as compared with that of the grooved
housing is illustrated in figure 11. For the grooved housing, when the cen-
trifuge action on the liquid becomes great enough, the gas bubbles are d i s -
placed to the rotor surface and pass over the lands. Therefore, the bub-
bles a r e not pumped to the high-pressure end. For the grooved rotor, the
centrifugal action on the liquid forces the gas bubbles into the grooves;
therefore, the bubbles are pumped to the high-pressure end of the seal.
When a viscoseal is ingesting gas, the axial pressure gradient is non-
linear and increases as the liquid-to-gas volume ratio increases in the di-
dj^ rection toward the high-pressure end. For no gas ingestion, the axial
pressure gradients a r e linear as shown in figure 8, which contains data
5.1.16

on sealing oil. As shown in figure 12, gas ingestion results in nonlinear |


pressure gradients (increasing slope with increasing liquid-gas volume),
which a r e significantly different from the gradients shown in figure 8.
Leak rate. - In both viscoseal combinations (grooved housing and
grooved rotor) water was sealed, with no detectable leakage. It was ob-
served that, although gas ingestion reduced the sealing capacity XG, ef-
fective sealing of the water was maintained.
Results of Sodium Sealing Studies
Grooved housing. - In sealing sodium, the grooved housing did not
ingest gas (fig. 13) for concentric alinement (0. 005 in. F. I. R.) over the
range investigated (Re* = 8. 66 to '50. 5) and for p r e s s u r e s ranging from
2 to 100 psig. The sensitivity of the leak detector to measure gas inges-
tion rate was estimated to be 0. 50 cubic centimeter per hour. It was ob-
served that nonconcentricity leads to gas ingestion, but this situation was
not investigated.
Grooved rotor. - The grooved rotor started to ingest gas at Re* = 10,
and the rate increased with increasing Reynolds number, as shown in fig-
ure 13. This ingestion characteristic is similar to that observed when
sealing water with the grooved rotor.
Gas ingestion into closed cavity. - In sealing sodium, gas ingestion
was readily observed by monitoring the liquid level in the leak detector.
As shown in figure 14, a straight line on the oscillograph paper indicated
no gas ingestion, no leakage, and thermal equilibrium. When gas inges-
tion occurred, the liquid level line showed an upward trend, the slope of
which is the ingestion rate. As gas ingestion proceeds, the gas bubble in ^p
the pressurized cavity increases in size (fig, 14(a)), and when the size
5.1.17

increases to the rotor outer diameter, the gas flows back out the annulus
formed by the rotor and housing. This causes a decrease in bubble size
and a sharp drop in the leak detector readout trace. If the interface is far
enough away from the seal low-pressure end, the liquid will be scavenged
back within the nonwetted length, and no loss of liquid will occur. The in-
gestion process will begin again and the net result is a saw-toothed trace,
each sharp drop indicating gas blowback. If the interface is sufficiently
close to the seal end, liquid loss will occur, and the leak detector readout
will show a saw-toothed trace with a downward trend, which represents
a liquid loss at each blowout (fig. 14(c)). This latter action may be the
"seal breakdown" phenomenon reported in reference 21 and the "secondary
leakage" phenomenon reported in reference 4. (In sealing water, the r e -
peating bubble growth and subsequent seal blowout was visually observed
by means of a transparent housing.)

Sealing capacity. - The grooved housing had a higher sealing capacity


(higher sealing parameter, AG) than the grooved rotor. The comparison
is given in figure 15, which gives the sealing parameter AG a s a function
of Reynolds number based on c (Re ). The difference between the value of
AG for the grooved housing and the value for the grooved rotor was at-
tributed to the difference in gas ingestion characteristics. The grooved
rotor operates with a gas-liquid mixture along the full wetted length simi-
lar to that shown in figures 9 (b) to (e); the grooved housing operates with
a homogeneous liquid at the high-pressure end of the seal and a gas-liquid
mixture at the low-pressure end similar to that shown in figure 9(f).
Figures 16(a) and (b) show the sealing parameter AG as a function of
Reynolds number (Re^) for groove configurations of five- and ten-helix
5.1.18

starts (see table II) with and without secondary grooves on the lands. In
both cases, a higher sealing coefficient was obtained through the use of
secondary grooves. Secondary grooves also operate at lower power ab-
sorption levels for equal sealing capacity, as was determined by the
amount of heating power required to maintain operating temperature.
This lower power absorption is due to less land area as compared with
the land area of the seals without secondary grooves.
Sodium was sealed-for p r e s s u r e s ranging from 2 to 100 psig with
negligible leakage loss. The operating range was within Reynolds numbers
(Re*) of 8. 66 to 50. 5, and the fluid temperature was in the range 300°
to 625° F. The maximum length of any one run was 8 hours and accumu-
lated testing time amounted to 50 hours for 10 runs. The oxygen content
of the sodium for the 10 runs varied between 50 to 95 parts per million,
as determined by the analytical method described in reference 31.
THEORETICAL ANALYSIS OF THE VISCOSEAL
Basic Model

The rectilinear Cartesian coordinate system was selected for mathe-


matical tractability. A parallel groove-ridge plate geometry with a
smooth flat plate moving relative to the parallel groove-ridge plate, both
of infinite horizontal extent, was chosen as the basic model (see fig, 17),
Conceptually this model can be thought of as an infinitely long smooth
rotor with a very large diameter and a concentric grooved housing. Actu-
ally this infinite grooved plate model is the limiting case of a helical
grooved cylinder where R - < » o r c —0. H c / R < 0 . 0 1 , this model
should give a good qualitative picture of the flow field and p r e s s u r e gener-
ation and significant trends for optimum geometry. (See argument in
5.1.19

appendix A.) For example, a practical geometry is a 2-inch-diameter


rotor with a 0. 005-inch concentric clearance, c/R = 0. 005.
The clearance between the infinite plates is filled with any incompres-
sible fluid, and it should be noted that for this infinite horizontal extent
case there a r e no end effects (in practical engineering design, of course,
end effects must be accounted for). Since experiments at Lewis show that
the end effect zone decreases with increasing rotor speed and also that it
is less than a groove width, this analysis should be valid in the bulk of the
homogeneous fluid (see figs. 18 and 19).
In the flat-plate case, the boundary conditions on a groove-ridge plate
in motion a r e constant; however, it will be advantageous to have the
smooth flat plate move relative to the grooved flat plate. This will elimi-
nate the temporal inertia t e r m s and reduce the subsequent independent
variables by one.
A word of caution for extension of the results to a case where the
groove-ridge pairs on a finite shaft a r e rotating relative to a fixed smooth
housing. The boundary conditions a r e different when the smooth housing
is stationary and the grooved rotor is moving, and when the housing is
grooved and the rotating shaft is smooth. When the rotor is grooved, the
surface speed is different at each point along the step or edge of the moving
boundary. Thus this analysis should be more valid for a smooth rotor and
grooved fixed housing (from ref. 5 the grooved housing appears to be the
desired mode of operation) with a large radius of curvature and/or small
clearance.

When referring to figure 17, it is noted that by the choice of coordinate


system the groove-ridge pairs appear only in the y, z-plane. Later it will
be shown that this choice of coordinate system will be necessary to reduce
the analysis to a two-dimensional problem for the quasi two-dimensional
flow field. It should also be noted that the drag force has been resolved
into its components in this coordinate system and there is no equilibrium
of forces in the z-direction with inertia (nonlinear) effects neglected.
BASIC EQUATIONS
The Navier-Stokes equations for a homogeneous, incompressible,
laminar, Newtonian fluid are the following (see ref. 32):
x-direction:

P°''=F, ^^J^^^^^\ (1)


Dt ^

y-direction:

P^^'K ?l^J^^^^^^] (2)


Dt y 8y W 8y2 3z2,
z-direction:

R,J^^^,^\ (3)
Dt ^

The Eulerian or spatial derivative is D/Dt, which is composed of the


local acceleration (temporal velocity change) and the convective accel-
eration (spatial velocity change):
D a s a „ a
= +U +v— +w
^*
I 9t J ^ ex 3y az^
Local, unsteady, or Convective
nonstationary accel- acceleration
oration
The incompressible continuity equation is
^ +^ +£E = 0 (4)
ax ay az
The incompressible energy equation with constants C and k and neglect-
ing compression work DP/Dt (ref, 32)

Dt W2 3y2 3z2y
where # is the dissipation function
.2 ,^,.2 ,^...21 + /9Z + /aw ^ avA2
*-p^©^ey lax
9UA2
ay/ \ay az/

+ f iH + ^ ^ . 2 /9u ^ av _j_ awY


.az ax/ 3 lax ay az/
The assumptions for this analysis are as follows:
(1) For steady flow,
au „ a v „ aw _= 0
at at at
For no body forces )

F x - ^y- Fz == 0

This means there are no electromagnetic fields present and gravitational


effects are negligible.
(3) Isothermal and equal temperature plate surface conditions are
assumed.
(4) Viscous (frictional) heating is negligible. Therefore, the thermo-
physical properties are constant. This appears to be a good assumption
for order 10 (water) and smaller (liquid metals) Prandtl number Pr fluids:
C u
Pr = -„£»
k
In experimental setups there is an inherent unsteadiness due to vibra-
tions, shaft runout, and eccentricity of rotor with respect to housing and
mlsalinement. These factors will influence the degree of correlation be-
tween the analytical and experimental results.
The conservation of momentum and mass equations were nondimen-
sionalized and a formal ordering "procedure was used to determine the
relative magnitude of the terms. Consequently, in a formal way the neg-
ligible terms were found (see appendix A).
An important parameter, the modified or reduced Reynolds number,
was found:
.2
R e * = R e L ( ^ ] = ^%{j) ^ ^/L < 1

This modified Reynolds number gives the relative magnitude of the inertia
forces to the viscous forces; that is,
Re* = I^srtia forces
Viscous forces
Thus, inertia forces can be neglected only if Re* < 0. 01.
After the formal ordering procedure (see appendix A), the three-
dimensional flow field equations are as follows:
5.1.23

h
Conservation of momentum:

u9H + v ^ + w ^ = - l i £ + . l ^ (x-direction) (6)


ax ay az p ax 9y^

IP^o (y-direction) (7)


ay

aw , ,, aw ,„ aw 1 ap ^ „ p ^ w ^ a^w' (z-direction) (8)


U + V— + w —
ax ay az p az vay^ az'
Conservation of mass:
_au av aw _. Q
(4)
ax ay az
Fully developed flow is assumed in the x-direction. This means that
V and aP/ax a r e not functions of x and that the pressure gradient in the
x-direction (parallel to the groove) is a constant (verified experimentally
in ref. 5). In figure 18, it is observed that this assumption is valid only
where the average p r e s s u r e profile at the ridge leading or trailing edges
in the axial direction is linear. Gas ingestion results in a nonlinear pro-
file (see figs. 18 and 19).
With the assumption of fully developed flow along the groove-ridge
direction (x-direction), the flow field equations become two-dimensional:

,V au + w
, au m+p^ ap = constant
where — (9)
ay az p ax ay 2 ax
•I
if s 0 (10)
ay

Vi^ + w i l = - l 9P+./l^H.i!^^ (11)


ay az p az \^^2 ^^2)

Conservation of m a s s :

i l +l ^ = 0 (12)
9y az
The boundary conditions for an arbitrary shaped groove-ridge geome-
try (see fig. 20) are:
(1)
v= 0
u = U cos a /at y = 0
w = -U sin a.
(2)
u =v =w =0 at y = h(z)
For example, when h(z) is a rectangular groove-ridge pair the boundary
condition is:
u =v =w=0 at c < y < c + h (on side of groove)
u = v = w = 0 at y = c (on ridge)
u = v = w = 0 at y = c + h (on groove root)
(3) Periodicity of groove-ridge pairs at the clearance interface.
For 0 < y < h(z)
v(0, y) =- v(L sin a,y)
w(0, y) ^ w(L sin a, y)
P(0,y) = P ( L s i n « , y ) - (AP^+g^z
where (AP J ^ = const
5.1.25

W
All velocity derivatives must also be the same to satisfy the periodicity
boundary conditions

a"+"^v(0, y) „ a"+"^v(L sin a, y)


ay" az«^ ay" a z "

a"+™w(0, y) ^ 8"+"^w(L sin a, y)


ay az ay az

for 0 < n, m < <»


(4) Boundary condition for BF/dx for a cylindrical geometry (visco-
seal).
In appendix B it is shown that for a cylindrical geometry BF/dx de-
pends on the (AP ) obtained in the z-direction

ap N(AP^_)
r+g'z
ax 27rR cos a
These equations (9), (11), and (12) can be solved for the velocity com-
ponents. The velocity components can be obtained for arbitrary (AP „) .
Then the axial pressure gradient dF/dV can be found from

ap ^Paxial ^N)(AP^ )
= = _2— cot a
dh' V L
where L ' is the axial length (see fig. 21). Qualitative pressure distribu-
tion from experimental data in the x - and z-directions is shown in figures
22 and 23.
By examining the analysis, boundary conditions, and the derived equa-
tions, the axial pressure gradient dF/dV is a function of (a, U, p, i^),
where cr is a shape parameter that is a function of (b/h , c / h , a/b, a).
5.1.26

4
Since convective inertia effects are important, new geometric parame-
ters appear which are not implied in creeping flow theory. Most notable of
these is the aspect ratio b/h , which implies the degree of convective ef-
fect. When the b/h and a/b ratios are known, N, the number of helix
or groove starts, can be found. (Direct effect of N is seen from boundary
condition (4).) Thus the number of groove or helix starts also is an impor-
tant parameter. Since inertia is important, a high density is desired for
pressure generation.
The equations (9) to (12) and boundary conditions are the minimum
number of terms that are necessary to mathematically describe the physics
of the internal flow. This is the simplified form of the Navier Stokes equa-
tions which should be used for an analytical solution; however, a mathe-
matical function to describe the groove-ridge shape may be too complex
for a closed-form analytical solution. Thus the equations are solvable by
using high speed digital computer numerical methods. Since numerical
schemes can accommodate all two independent variable terms with equal
ease, it is suggested that the complete set of two-dimensional equations be
solved numerically, i. e.,

v i I + w l Z = - l l P + . f l ! z + l?Z^ (13)
ay az p ay ^^^2 3^2j

vi^ + w i l = - l i P + . / l ! ^ + i M (11)
ay az p az \^^2 ^^2)

viH + w i H = , . l 9P+ JI!H+9!H\ where !£ = constant (14)


ay az p ax \^^2 ^^2) ax

11 + ^ ^ = 0 (12)
ay az
5.1

Of course, the numerical solution will determine under what condi-


tions the previously derived simplified form of the two dimensional Navier
Stokes equations are good approximations.
By examining the set of equations (11) to (14) it is seen that v and w
can be calculated from equations (11), (12), and (13). The solution is
identical to a two-dimensional flow case. Then the u flow field can be
found from equation (14) which is linear in u. This is an example of the
Independence Principle.
From appendix A the energy equation with no temperature variation
along the groove-ridge direction (BT/dx ~ 0) is

^!l,^!l.J.[^.fl],JL^,^ (15)
ay az Pr^gy2 g^V PSC pgC
Comparing the above equation with the x-direction momentum equation
it is seen that the mathematical form of both equations is identical.
Substituting 6 =pu
and

-iL * o r ^ - a constant ::»- l 1 1


pgC pgC p ax
Thus with P r - 1 and the above substitutions, the equations (14)
and (15) a r e identical. The thermal boundary conditions are
At y = 0
6 ~ constant =^w =~ _u sin a = constant
At y = h(z)
0 -.- 0 => w -= 0
and the boundary conditions a r e similar in form.
The above statements mean that once the u velocity field is found
the solution will also give the temperature distribution for the stated r e -
strictions of constant frictional heating or a heat source in the fluid with
the appropriate physical parameters equivalent to the constant p r e s s u r e
gradient along the groove-ridge pair direction and P r = 1. Note that this
temperature distribution solution violates the initial assumption of con-
stant thermophysical properties; however, for small temperature dlffer-
ences the above solution will give a good qualitative picture of thermal
effects.
DBCUSSION
Discussion of Equations
This analysis is valid only fo-r a homogeneous fluid and figures 18
and 19 clarify the region of validity. The analysis is not valid for a
scavenging area and a region consisting of a mixture of gas and liquid
(see fig. 18) or under conditions of gas ingestion.
It should also be pointed out that the analysis is not valid in the end
effect region at the high p r e s s u r e end of the seal (see fig. 18) where the
groove connects with the cavity. Since the appropriate equations a r e non-
linear, the p r e s s u r e is not a harmonic function; thus, the usual analog
methods of finding end effect a r e not available as found in creeping flow
regions (see appendix C).
The equations a r e unique in many ways. Mathematically, the problem
has been reduced to two dimensions (y- and z-directions) but all three ve-
locity components a r e present. Thus the flow field can be described as
quasi-two-dimensional. Also, there is a pressure gradient in the third
dimension (x-direction), but fortunately this gradient is a constant and
dependent on the y- and z-direction gradients.
The velocity component v is no longer negligible as it was in the
creeping flow solution. Physically, v cannot be negligible because of the
presence of the edge effect which will result in a component of velocity in
the y-direction.
Since there is no restriction on net volume flow, the pressure gen-
eration prediction equations should equally apply to hydrodynamic seals,
bearings, and pumps.
When further examining the physics of the problem, it is seen that
across the groove-ridge pairs the unbalance of the convective accelera-
tion force with the drag force results in a pressure gradient in the
z-direction. Since the step or edge causes the convective acceleration,
its effects a r e not negligible unless there is a very large aspect ratio
and the mathematical model accounts for the edge effect. Along the
groove and ridge (x-direction) the drag force is in equilibrium with the
induced pressure and convective inertia force. The drag force not only
causes fluid motion but also aids in maintaining equilibrium with the
p r e s s u r e along the groove or ridge.
Both equations of motion illustrate the coupling of u, v, and w in
the convective inertia t e r m s . This means that physically a condition such
as no end leakage is possible. In creeping flow there is no coupling be-
tween the velocity components, which illustrates the shortcoming of that
analysis ( i . e . , end leakage is not zero).
The assumption that the inertia t e r m s are of the same order as the
viscous t e r m s results in equations of the boundary layer type; however,
the boundary conditions a r e different.
5.1.30

4
The equations a r e of a form that might tempt one to reduce the num-
ber of independent variables by one by trying to find the admissable flow
field which would result in a similarity transform solution. This tech-
nique fails because of the characteristic lengths that a r e present in this
type of confined flow.
The equations a r e solvable by using high-speed digital computer nu-
merical methods. A mathematical function to describe the groove-ridge
shape may be too complex for a closed-form analytical solution. It i
should be interesting to find out if the step is still the optimum shape
with convective inertia effects included. For creeping flow, Lord Rayleigh
found in 1912 that the step was the optimum shape by using calculus of vari-
ations.
The equations and ordering procedure should be reexamined carefully
before extending these results beyond the assumptions used in this analysis.
Discussion of Modified Reynolds Number
A check was made on the significance of the modified or reduced
Reynolds number by calculating Re* from the available published data at
the claimed onset of turbulence (breakpoint). In table III it can be noted
that the departure in all cases from a constant sealing coefficient occurred
when Re* was in the range of 0. 26 to 2. 2, which suggests that the so-
called beginning of turbulent operation is really the beginning of significant
convective inertia effects. This would be in agreement with Kettleborough
(ref. 29), who found qualitative agreement between his inertia solution and
experimental turbulent results.
For seals 5, 6, and 7 (see ref. 27 and table III), only the groove to |n
ridge ratio has changed, As the ratio increases, the breakpoint decreases.
5.1

This is reasonable since a larger ridge area results in a less influential


convective inertia effect and delays the beginning of pronounced convective
inertia effects. It can also be seen in table III that the calculated Re* are
within an order of magnitude of one another for all of the experimenters.
It is difficult, nevertheless, to select the breakpoint (see fig. 24). Even in
a range of possible breakpoint numbers, however, the order of magnitude
will remain the same. Further experimental results will be necessary, but
in general, it can be concluded that if Re* > 1, a constant sealing coeffi-
cient or bearing number will not exist.
In figure 25 it is seen that classically the empirical sealing parameter
AG was thought to be a constant until some critical Re and then to vary
exponentially in the turbulent regime. The creeping flow solution was
thought to be valid until the critical Re and then a semiempirical turbulent
theory had to be used. Actually, the onset of significant convective inertia
may be gradual, as shown in figure 25 Data (from ref. 27) for seal 7 in
figure 24 tends to show this. The derived equations herein pertain to all
cases under assumptions stated and will break down only under strong turbu-
lence or an unknown instability phenomena. However, even in the turbulent
flow model the convective inertia t e r m s must be considered in addition to
the Reynolds* s t r e s s t e r m s . This is specifically shown by Kulinski and
Ostrach (ref. 33) in their critical evaluation of high speed fluid film lubrica-
tion theory.
Principle of Viscoseal Operation
The classical explanation of viscoseal operation is now described.
Since the seal is "pumping" the sealed fluid as fast it is leaking out, there
is no net end flow (zero leakage). From the basic laws of fluid mechanics,
however, it is seen that the viscoseal axial pressure gradient results from
the unbalance of the convective inertia force across the groove-ridge pairs
and the drag force. The convective inertia force is much greater than the
opposing drag force, and this results in a net pressure force. Along the
groove-ridge direction, the drag force maintains equilibrium with the p r e s -
sure force and convective inertia force. All three velocity components a r e
"coupled, " and thus zero net leakage is achieved. This is an excellent ex-
ample of the physical importance of a nonlinear effect. The inadequacy of
the linear theory (creeping flow) has been previously discussed.
SUMMARY OF RESULTS
P r e s s u r e Patterns
P r e s s u r e patterns were studied for a viscoseal in the laminar-flow
regime with a mineral oil and the following results were obtained:
1. For the condition of helical grooves in direct connection with the
pressurized cavity, the end effect is evidenced by a sharp decay of land
leading-edge p r e s s u r e .
2. No significant change in end-effect length was noted over the speed
range investigated (modified Reynolds numbers (Re*) of 0. 04 to 0. 25).
3. Circumferential pressure gradients in the plane of rotation (plane
orthogonal to ?]-axis) increase linearly across the groove and decrease
linearly a c r o s s the land. This p r e s s u r e pattern repeats for each groove-
land pair and provides a saw-tooth pressure profile around the circumfer-
ence. The pressure gradients along the groove and land were linear.
Gas Ingestion and Sealing Capacity
Visual observations and experimental evaluation of viscoseal a s s e m -
blies with water (modified Reynolds numbers of 1. 75 to 10. 84) and sodium
5.1.33

(modified Reynolds numbers of 8. 66 to 50. 50) as sealed fluids disclosed


the following:
1. Gas ingestion characteristics were fundamentally different for the
helically grooved rotor-smooth bore housing combination. In the grooved
housing-smooth rotor combination, centrifugal action tended to hold the
gas bubbles out of the groove; thus the bubbles tended to escape over the
lands rather than be pumped into the seal high-pressure end and a condi-
tion of no gas ingestion could be achieved. In the grooved rotor-smooth
bore housing combination, centrifugal action forced the gas bubbles into
the rotor grooves; thus the gas bubbles were pumped into the seal high-
pressure end and increasing speeds resulted in increasing gas ingestion.
2. In sealing sodium, the grooved housing-smooth rotor combination
had no detectable gas ingestion rate, but the grooved rotor-smooth bore
housing combination showed increasing gas ingestion with increasing
Reynolds number.
3. No gas ingestion occurred when sealing water with either the
grooved rotor-smooth bore housing or the grooved housing-smooth rotor
viscoseal when the liquid interface was stable (below Reynolds number (Re*)
of 1.9).
4. In sealing sodium, the grooved housing-smooth rotor combination
produced a higher sealing capacity (higher sealing parameter) than the
grooved rotor-smooth bore housing combination. This was attributed to
the presence of gas ingestion in the case of the grooved rotor.
5. Sealing, with negligible sodium leakage rate, was obtained for
short operational periods (maximum length of continuous evaluation was
8 hr) over a pressurB^range of 20 to 100 lb force/sq in. gage and at sodium
temperatures of 300*^ to 625° F.
Theoretical Analysis of the Viscoseal
A review of the literature on the theoretical analysis of the viscoseal
revealed that either the convective inertia forces or the viscous forces
were neglected (inviscid fluid assumption). This analysis included both
convective inertia and viscous forces. The following two-dimensional set
of equations for a quasi-two-dimensional flow field were formulated and
a r e solvable for the velocity and p r e s s u r e distribution and optimum geom-
etry utilizing numerical methods on a high-speed digital computer:
Conservation of momentum:

v i H + w ^ : = - l ^ + i ^ ^ where ^ = constant (9)


ay 8z p 92 ay2 ax

(10)
ay

aw w 9w _ 1 ap ^ J^a^w ^ a V (11)
V +
ay az p az 'Vay2 az2

Conservation of mass:

3v ^ aw _ Q (12)
ay az
With the following boundary conditions
(1) At the moving smooth plate surface,
u = U cos a V=0 w = -V sin a
(2) At the fixed parallel groove geometry surface,
u -= V = w = 0
5.1.35

(3) Periodicity of groove-ridge pairs at the clearance interface.


For 0 < y < h(z)

v(0,y)^v(Lsin«,y) and ^1^1^^ = ^I^IML^^L^^


ay" az™ ay" az™

w(0,y)=w(Lsinc,y) and n+m,„/^ „\ ._ ^-'-ML


^"^"'^0,7) -,n+m. sin a,y)
ay az 3y dz
P(0, y) = P(L sin a, y) - i^P^^g)^ for 0 < n, m < *

(4)
ap ^ N(APr^g)^
ax 27rR cos a
From the development and analysis of the equations, the following
results and conclusions can be made:
1. A modified or reduced Reynolds number evolved which gives the
relative magnitude of the convective inertia forces to the viscous forces:

Re* = Re-r (^\ = Re (^\ if 1 < 1

The characteristic length was selected to be the length of the groove-


ridge pair in the direction of the smooth plate velocity vector. This is
the principal length over which the driving force acts. The flow behavior
over each groove-ridge pair is the same. Convective inertia effects can
only be neglected if Re* < < 1, which will result in true creeping flow.
2. A modified Reynolds number equal to 1 will predict a significant
change in sealing coefficient or bearing number which heretofore was at-
tributed to the onset of turbulence This means that the convective inertia
effects a r e the foremost contributing factor to the improvement in sealing
coefficient or bearing number, not the turbulence effects. A calculation
of the modified Reynolds number at the onset of turbulence of four different
experimenters disclosed that the modified Reynolds number was near 1
(Re* = 0. 25 to 2. 2). Thus a turbulent flow analysis must also include the
convective inertia terms in addition to the appropriate Reynolds stress
terms.
3. Several additional parameters of pressure generation dependence
were found that are not implied in the creeping flow analysis. These addi-
tional parameters, which arise from convective inertia effects, are the
fluid density p, the groove width to depth or aspect ratio b/h , and the
number of groove-ridge pairs N. .
4. Solving equations (11) to (14) by numerical analysis will permit the
investigation of arbitrary groove-ridge shape, for example, sinusoidal,
rectangular, triangular, etc. Now a check can be made to see if a step is
the optimum shape with convective inertia effects as it is for creeping flow.
5. The analysis provides a unified approach to parallel groove opera-
tion previously covered by creeping flow (sometimes called laminar flow)
analysis and a semiempirical turbulent flow analysis.
6. No restriction was placed on the net flow situation between the
plates. Therefore, the analysis can be used to analyze parallel groove
geometries in hydrodynamic seals, bearings, and pumps.
7. The viscoseal axial pressure gradient results from the unbalance
of the convective inertia force across the groove-ridge pairs and the drag
force. Along the grooves or ridges, the drag force maintains equilibrium
with the pressure force and the convective inertia force.
5.1

8. Zero leakage in viscoseal analysis can be attained through the con-


sideration of convective acceleration terms (nonlinear effect). It is argued
that the creeping flow analysis, which is a linear theory, cannot prove zero
leakage.
9. The format of this analysis can be used to evaluate qualitative con-
vective inertia effects in any step or converging wall hydrodynamic seal,
pump, or bearing (Rayleigh step, journal bearing, etc.).
10. Solution of the^u velocity (along the groove-ridge direction) will
give an insight to the heat transfer in the viscoseal for the case when
Pr = 1.
APPENDIX A
FORMAL ORDERING PROCEDURE FOR SIMPLIFYING
BASIC EQUATIONS
The Navier-Stokes equations for a homogeneous incompressible
Newtonian fluid, neglecting body forces and for steady flow, are

p(u9H + v ^ + w ^ U - ^ + M f - ^ + - ^ + ^ ' l (Al)


ax ay azi ax \^^2 ^^2 9^2,

pfu^+vll.w^ZV-^+Mf^ +^ +^ l (A2)
ax ay az/ ay \^j ^^2 9^2^

\ ax ay azi az \^j ^^2 ^^2J

The incompressible continuity equation is


au av aw n /Ay|\
— + — + ™_ = 0 (A4)
9x ay az
To find the important terms in the previous flow field equations, a
formal ordering procedure will be used to determine the relative magni-
tudes of the t e r m s .
The t e r m s a r e normalized by scaling them to their appropriate maxi-
mum characteristic value as follows: let

L cos ff c L sin a

U COS a V U sin a
The characteristic "dimensions c and L can be seen in figure 17.
It is noted that a trigonometric relation exists between x and z charac-
teristic lengths and that c/L < 1. The characteristic length was selected
to be the length of the groove-ridge pair in the direction of the smooth
plate velocity vector. This is the principal length over which the driving
force acts. The flow behavior over each groove-ridge pair is the same.
Let p* and v* be the fluid density and viscosity corresponding to a
reference state (P*, T*). Therefore,

17=ii=l

and
p = £^=.i
P*
for incompressible fluids with constant therm ©physical properties. Now
to find the relative magnitude of V so that all t e r m s of the incompressible
continuity equation will be of the same order, the incompressible continuity
equation is nondimensionalized:
5.1.39

b
To have all t e r m s of the equation the same order of magnitude, —— must
_ , U c
be of the order of 1. Therefore, V = U — , and thus v = — —.
L U c
Next the equations of motion are nondimensionalized. The pressure
term must first be nondimensionalized; even though the pressure is not
the driving force, it is the function of primary importance. Thus the char-
acteristic p r e s s u r e is not extremely important. Since the pressure gradient
is related to the rotor speed, it appears to be logical to nondimensionalize
the p r e s s u r e with respect to the dynamic pressure pU (also called the
dynamic head):

PU2
It can be seen that the characteristic dynamic pressure pU has r e -
duced the number of nondimensional groups by one (i. e., if P * (a refer-
ence pressure) had been used, another nondimensional group would be
present in the equations). Also, since the incompressible case is being
analyzed, the p r e s s u r e appears in the dynamic role only, not in the thermo-
dynamic role.
Nondimensionalization of x-Direction Momentum Equation
When the x-direction momentum equation is nondimensionalized,
equation (1) results in

u _31 au + w----==
- + v-— du 1 ap

ax ay az „^„2 ax
•^ cos a
I7* 1 d\ ^ (V^ d\ ^ 1 d\
+ —
UL
Next a Reynolds number is defined based on the characteristic length L:

Re, =^

where
p = p*'v = V*

and thus
17 au . „ au , =. 31 1 ap"
U +v .+ W — = - —
3x a? dz „^„2 ax
' cos a

+ ^ I ^ a ^ ^ fhy 2 -d^
^ ^ 1 a\ (A6)
Re^ ^2„. .~2
— \c J ^--2
— 2 —2
cos^ff ax^ ^"^ ' ay"" sin a 3z
The convective acceleration and pressure t e r m s are of unit order; thus,
the viscous term must be of unit order also. The largest term of the
22 -.2^
2— ,.. , ,xv2
viscous forces is (^] •^-^, since (—] < 1; therefore, the _ i — ( _
\c / p-»2 VLJ Re, \ c ,
ar ^"' ""L '^ ^
should be of unit order. The modified Reynolds number is defined by
Re* = R e L l ^ I • This is the identical "reduced" or "effective" Reynolds

number that is used in the bearing theory to show the true relative magnl
tude of the inertia and viscous forces; therefore,
Rp* = Ifl^rtia forces
Viscous forces
(e. g., see ref. 32). Now for curved geometries, a Reynolds number
based on clearance has appeared. This is especially true in viscoseal
analysis where
5,1.41

Writing the previous x-direction momentum equation in t e r m s of Re


gives
^ 31 ^ 3 1 ~= du 1 a p , i^*/c 1 d\^h d\ c 1 aV
u __ + v-—+ W —= ^ —+— — _+— _ + _
3x 3y
^ az ^^„2^
cos a ax UclL \ ^„^2^
cos a ^-2
ox c ox
„ 2 L „.
sm 2^a dz
^^2J
or

^ 3u_._^, 3 1 , ^ ai„^ 1 ap^ 1 fc 1 a\_^L a\^.c i aV


31 3? ^ az „^^2
cos a ax Re^lL
c \ „^„2^
cos a -,^2
ox c -,-^2
oj L „.
sm 2^a ^^2
9z '
2— —2
Again it is seen that the a u/9y term is the largest.
Thus 1/Re (L/c) should be of unit order. Now

nL
which leads to
2
R e * = RCj^
(i) - M I )
It does not matter which Reynolds number is used to find the relative
inertia force to viscous force ratio provided that c/L < 1, which c o r r e -
sponds to the conditions
e2 »
a2
3y2 3z2

and
32 32
»

3y^ 3x2

The proper x-direction momentum equation then becomes

U9H+V^ + W^==^1 9P+P1!H (A7)


ax 9y az p ax g 2
It is assumed that a does not go to 0° or 90° for the formal ordering
procedure to be valid.
Example
A numerical example will be used to illustrate the validity of neglect-
2 2 2 2
ing the 3 /3z and 3 / a x viscous t e r m s . A typical viscoseal has the
following construction and operating conditions: smooth 2-inch-diameter
rotor, 0. 0035-inch concentric radial clearance between the ridge (housing)
and rotor surfaces, 1000 rpm, and sealed fluid (water) at 60° F. There-
fore,
V - 17.5X10" sq i n . / s e c
c = 0. 0035 in.
U= 27TR(1000 rpm)/60 = 100 i n . / s e c (plate speed)
c/R = 0.0035 (therefore, curvature effects should be slight)
L = Ridge width + Groove width (in circumferential direction)
= 0.2 in. + 0 . 2 in. = 0 . 4 in.
a= 15°
R e ^ = UL/^ = 23, 000
Re^ = Uc/i/ = 200
Using the Reynolds number based on the characteristic length L yields
I d%
,2^
^ // TLy a^
\2 ,2- . 32-~
Re^ .2„. ,^2
— \c J ^^2
__ 12 a—2u
cos'^a 3x^ "^ ' dy^ sin « 3z

1. i i ~ H + 13 0 0 0 l ^ + m ^
^«L V 3x2 9^2 9^2

Using the Reynolds number based on the clearance c gives


5.1.43

h
1 d\ ^ fh\ d\ ^_ /c 1 d\
Re. (i)z: 2 —2
COS « 9x
c / 3^2
ay
L /' sm
. 2a az
5-2 __

= A . 4. 011?1 + 114 .!^ + 0. 13 1 ^ '


^®c \ dx^ 3y2 3z^

l i both equations it is seen that d\/dy is the largest term. Now the
modified or reduced Reynolds number is
2
Re*-=ReL^|] =Re^^|^=L75

Nondimensionalizatioii of y-Direction Momentum Equation


Nondimensionalizing the y-direction momentum equation (eq. (2))
yields

l ^ + f 9 v ^ ^ 3 v _ / L X2 - .3^ P ^ ^ 1 3^
ax 3f az \c / 3y Re cos^a dx^ ^^ ' 3y^ sin^a 3z^J
(A8)
Again using the definition of Re* and examining the terms of unit order

u^, v ^ , w l Z , 1 8 ^ mi)
ax' ay' az' Re* g™2

thus
2 n^
0(1)
\c / ay

Since

ay

therefore
^ s 0 (A9)
ay

This result is expected for a flat plate where there is no mechanism to


generate a radial pressure gradient (e. g,, a centrifugal force). For
small curvatures, c/R < 0. 1 however, the radial pressure gradient
3P/3y should also be a small quantity and have little physical influence
on the axial pressure gradient. This can be shown by considering the
following simplified model. Consider the flat plate model accelerating
steadily about an origin. From elementary dynamics
a = t . § . + a„§„
t t n n
where: a. = q = 0 for steady acceleration
2 / . „ ... . . . . ...2
= S_ = fa cos a - w sin a}'
a..
^ R R
q is the fluid velocity in the tangential plane (x-z plane) and
q = u cos a - w sin a (see fig. 11). §^ is the unit normal that is directed
towards the axis of rotation. This is the y-coordinate direction in the flat
plate model. Thus the centrifugal inertia force effects the y-momentum
equation only.
Add pa to the inertia force in equation (A2) and nondimensionalize
as before and let U= UU and R - RR. Thus equation (A8) becomes

l ^ +^ i l +w ^ + 4 f ^ f / l V ^ B i n « w + cos«l)2=^/Lf al
ax ay 3z R \ C / \ R / \ c / ay

1 1 dh^ fh\^ d\ ^ 1 d\ (A8a)


Re^ cos 2a ax2 ^^ ' ay2 sin2« Bz^
5.1.45

b Now
2
±. fL\ f±\ (-sin aw + cos au)^ -' O'H)

Comparing the largest viscous term, pressure and centrifugal inertia term

9 l = J L l ! l + ± /£.^ (-sin «w + cos «u)2


ay R e ^ 9y2 R \ R /

Now use a typical Re, (as in example) = 103 and apply restriction

- < 0.01
R-

9P ^ (10-3) ^ ^(l^i^sin «w f cos au)^


9y ay2 R

Again
9P^0
ay

if

^ < 0.01
R~~
which substantiates the claim that the centrifugal effects should be small.
In this model Coriolis forces were also thought to be negligible. The
above centrifugal effect is for the case when the grooves a r e on the sta-
tionary housing. A slight modification is required for the groove-ridge
pattern rotating. Note that as c/R — 1 the centrifugal force effects are
no longer negligible.
Nondimensionalization of z-Direction Momentum Equation
In a like manner the z-direction momentum equation is nondimension-
alized and appears in a similar form as the nondimensional x-direction
momentum equation (A6)
U ! H + v ^ ^ .+
-^w^ ^^-^
' ^ -= 1 3P_^^ V* 3^,, (hf ^+_L 3^
Sx 3y" 3z sm^a 9^ UL 2 —2 3y^ siE^ff 3z 2
cos ff ax
(A 10)
Examining the viscous t e r m s
.2 . ^
1 1 )V
Re, ^^ ^ 3y2 sin2« a^2_
cos 2 « 3 x 2
The largest term is (L/c)^ ( 3 ^ / 3 y ^ ) and it would appear that the 3 ^ / 3 1 ^
term could be neglected; however, this term must be retained due to a
physical argument. Because of the "edge" there is a sharp gradient in the
z-direction and the nature of the flow is such that the 3 w/az term must
be retained. Thus the z-direction momentum equation in dimensional form is

1 ap _^ „/3^w ^ d \
u i Z + v^E + w ^ (All)
3x ay az p dz
.3y 3z
This was pointed out by Dr. A. Mager of Aerospace Corporation, Los
Angeles, California (personal communication, October 5, 1966).
Nondimensionalization of the Energy Equation
The energy equation for steady temperature variation for an incom -
pressible fluid with constant thermophysical properties and negligible
compression work is
3T 3T „. 3T k /3^T 3^T 3^T\ M * Q*
(A12)
U —=- + V -^-^ + w
3x 3y 3z pgC \ 9^2 pgC pgC
3y az
Nondimensionalize as before and nondimensionalize the temperature by
letting

e= and # ' = —
'M
5.1.47

k Thus

_ 30 - 30 . ^ 30 3^0
u +V +w
ax ay 3z IpgC Uc' _ „ 22 a^ ax
VL] cos ,^2
—2 ,^2
-^
3y U i .^J
sin^a 3z^

#^
Ql
LpgC(TM V PgC(T^-"Tg)

now
/ k J i L\ J ^ _ J _ / L\ ^ 1 ^ J___
I pgC y Uc cy P r Re^ I c / PrRe* P*

where P *e is the modified Peclet number

or

^ aa
ax
- 30
u— + Y — + w — = —^
3y
•'
— 30
az
1

P*
e
(tP 3^0 ^ 3^0 + i £

cos 2 a 3x
„~2 3^2
3y ^^
1

sin^a 3z^
3^0

#' Q*
LpgC(T^-Tg) pgC(T^^Tg)

Now assume constant temperature in the x-direction (along the groove


ridge pairs) then

^ a a ^ ™ 30 3^0 2,2r #' Q*


V -— + w + +
ay 3z P _3y i2«^Li
sin'^a az LpgCCT^^Tg) ^gC(T^-Tg)

Now dimensionalize all t e r m s but the temperature

..
V— 30 +
^w,„ 30
— k ^ll + ^]+JL.^ + ^
3y 3z pgC \^gy2 3z 2 / pgC pgC
5.1.48

4
P r = CM
k
therefore

vi£ + wl^ = iL/l?^ + ^ V - ^ * + ^ (A13)


3y 3z Pr^9y2 ^^2J PgC pgC
APPENDIX B
FORMULATION OF ALONG GROOVE-Rn)GE BOUNDARY CONDITION
FOR A CYLINDRICAL GEOMETRY (VISCOSEAL)
A special case of parallel groove geometry is a cylindrical geometry
which is the helically grooved fluid film seal (viscoseal). For this cylin-
drical geometry a special boundary condition must be placed on the p r e s -
sure gradient along the groove-ridge direction. This can best be illus-
trated by referring to figure 26. It is seen that the point o to point a can
be traversed along both coordinate axes on the surface (x and z axes).
From figure 26 and figure 21, it is seen that going from point o to
point a along the z-axis involves the traversing of an angle cp.

tana=I^ (Bl)
L
Now going the same distance along the x-axis

cot a = 2 ^ ^ ^ (B2)
L'
Equating (Bl) and (B2)

tan a cot a

or
(p = 2v cos^a (B3)
The arc distance (path length) in the x-direction is:

S = YCR^)^ + (L')2
Upon substitution for L' and (p
S = 27rR cos a (B4)
The pressure differential at a from traversing the z-direction from
0 is

<^Po-a ^t-^Pr+g'zW
then
3P __ N(^Pr+g)z , N ( A P „ g ) , ^^^^
3x S 27rR cos a
It is seen that aP/ax = constant which follows from fully developed
flow along the groove-ridge direction and verified by experiment in refer-
ence 5.
APPENDIX C
SUMMARY OF CREEPING FLOW ANALYSIS
The optimization from creeping flow analysis has been formulated by
considering a flat-plate model (unwrapped cylinders, see fig. 27) and the
following:
(1) General incompressible plane Couette flow is assumed in the
^-direction (Poiseuille flow and simple Couette flow). The |-direction
momentum equation is

9 \ 1 ap
9^2 II d^

with the boundary conditions


u^ = U at C = 0
u^ = 0 at C = h
where
h = c at ridge surface
h=c+h at groove root surface
The well known solution (e. g., see ref. 32) is

^ h 2/i d | V h/
(2) Poiseuille flow is assumed in the TJ-direction. The 17-direction
momentum equation is

^ \ „ 1 3P
g^2 II df]

with the boundary conditions


w =0 at C = 0
n
w?] = 0 at C = h
The well-known solution (see ref. 32) is

w =A(l£V(?~h)
^ 2iU V3i7 /
Now the condition Q = 0 is imposed in the control volume; that is,
Q - Q^ + Q^ = f f \ d7] d? + ffw^ d? d | = 0

which means that there is no net end leakage or closed channel flow. This
is a restriction that limits the optimization and p r e s s u r e gradient predic-
tion to a seal only.
Solving for p r e s s u r e gradient that satisfies the Reynolds equation
(ref. 34) results in a sealing parameter
5.I.51

AP ^ 6jUUG

or
G == <=^ ' ^ p
G
GliVW / A
where A is the bearing number (ref. 34) and G is a geometry factor.
See references 1 to 26 for various expressions and values for G.
The creeping flow analysis has some paradoxes. It can best be illus-
trated by referring to figures 17 and 28 and by observing the resolution of
the drag force into components along the groove (x-direction) and normal
to the groove (z-direction). In creeping flow, the drag force is in equilib-
rium with the pressure force. Thus along the groove (x-direction) the
drag force is indeed in equilibrium with the pressure force. In the normal
groove direction (z-direction), however, the pressure and drag forces are
in the same direction, which is also the direction of a very large pressure
gradient (see figs. 22 and 23). Therefore, an unbalance of forces exists
in this direction. Physically, the Q = 0 restriction is not possible. Also,
this point can be illustrated from ^~ and rj-direction momentum equations,
respectively: ^2
^ \ _ 1 ap
9^2 M ai

' \ „ 1 ap
9^2 M a?]

Notice that there is no velocity "coupling" between the velocity compo-


nents; that is, u and w are independent of one another. It is obvious
that for zero leakage these velocity components are not independent of one
5.1.52

another. Thus, physically as well as mathematically it is not possible


under the restrictions of the classical creeping flow analysis to get zero
I
leakage.
Also in the creeping flow analysis the edge effects are assumed neg-
ligible. The full meaning of this assumption has not really been clarified;
it is only close to physical reality when c is very small and/or b/h is
very large. For example, the creeping flow analysis is valid for spiral
groove thrust bearings where c is on the order of 500 microinches and
the aspect ratio is on the order of 1000 (c/L « 1, see fig. 29). As
b — h , the edge effect becomes more pronounced as the convective forces
becomes more and more important. To date only reference 35 has ex-
plicitly stated this restriction.

To resolve the paradox of having the drag and pressure forces un-
balanced in the z-direction, a convective inertia force is in equilibrium
with the pressure and drag forces (a nonlinear effect). The step is causing
a convective change that results in generation of an axial p r e s s u r e gradient.
The role of the step is now described.
If a concentric rotor is rotating about a stationary smooth sleeve, no
axial pressure gradient is generated. This is the classic axisymmetric
Couette flow (see ref. 36). Furthermore, if an axial pressure gradient is
imposed ( e . g . , by having the rotor translate in the axial direction) an
"uncoupled" helical flow will result. Subsequently, of course, there will
be net flow out.
In the helical groove seal (viscoseal), the step or edge h is the
pressure generation mechanism. Without the step, that is, the limiting
case when h — o (rotor and housing smooth), there is no axial p r e s s u r e
5.1

gradient generated. Eccentricity will generate a radial pressure gradient.


Therefore, the effect of the groove wall is a convective accleration or in-
ertia effect (nonlinear effect). To neglect the convective force t e r m s
means that the primary axial pressure generating mechanism is neglected.
The creeping flow solution equations are linear and pressure is a po-
tential function. The creeping motion or slow viscous flow momentum
equation in vector notation is
VP = juV^?
The incompressible continuity equation says
V. V = 0
Taking the divergence of both sides of the equation yields

V • VP = /iV^(V . V) = 0
or
V^P - 0
which is Laplace's equation, since pressure is a scalar quantity. Conse-
quently, the multitude of mathematical analog methods of solving linear
equations and specifically Laplace's equation can be used. Thus, solu-
tions for end effect can readily be found (refs. 1 and 34).
Muijderman (ref. 1) used conformal mapping to analyze the spiral
grooved bearing. Again, this technique can be readily applied because
pressure is a potential function.
REFERENCES
1. Muijderman, E. A.: Spiral Groove Bearings. Philips Res. Rept.
Suppl. 1964, no. 2 . : See also Scientific Lubr., vol. 17, no. 1,
Jan. 1965, pp. 12-17.
5.1.54

4
2. Rowell, H. S.; and Finlayson, D . : Screw Viscosity Pumps. Engi-
neering, vol. 114, Nov. 1922, pp. 606-607.
3. Boon, E. F . ; and Tal, S. E . : Hydrodynamic Seal for Rotating Shafts.
DEG. Inf. Ser. 13, United Kingdom Atomic Energy Authority, 1961.
4. King, Alan E . : Screw Type Shaft Seals for Potassium Lubricated
Generators. IEEE Trans, on Aerospace, vol. AS-3, Supplement,
June 1965, pp. 471-479.
5. Ludwig, Lawrence P . ; Strom, Thomas N . ; and Allen, Gordon P . :
Gas Ingestion and Sealing Capacity of Helical Groove Fluid Film
Seal (Viscoseal) Using Sodium and Water as Sealed Fluids. NASA
TND-3348, 1966.
6. Ludwig, Lawrence P . ; Strom, Thomas N . ; and Allen, Gordon P . :
Experimental Study of End Effect and P r e s s u r e Patterns in Helical
Groove Fluid Film Seal (Viscoseal). NASA TN D-3096, 1965.
7. Vohr, J. H.; and Chow, C. Y.: Characteristics of Herringbone-
Grooved Gas-Lubricated Journal Bearings. J. Basic Eng., vol. 87,
no. 3, Sept. 1965, pp. 568-578.
8. Billet, A. B . : Hydraulic Sealing in Space Environments. Proceedings
of the Second International Conference on Fluid Sealing, B. S. Nau,
H. S. Stephens, and D. E. TurnbuU, e d s . , British Hydromechanics
Research Association, Harlow, Essex, England, 1964, pp. C2-17 -
C2-36.
9. Rowell, H. S.; and Finlayson, D. : Screw Viscosity Pumps. Engi-
neering, vol. 126, Aug. 1928, pp. 249-250.
5.1.55

h
10. Rowell, H. S.; and Finlayson, D . : Screw Viscosity Pumps. Engi-
neering, vol. 126, Sept. 1928, pp. 385-387.
11. Rogowsky, Z . : Mechanical Principles of the Screw Extrusion Ma-
chine. Engineering, vol. 162, no. 4213, Oct. 11, 1946, pp. 358-
360.
12. Strub, R. A . : Spindle Drag Pump. Machine Design, vol. 25, July
1953, pp. 149-151.
13. Pigott, W. T . : P r e s s u r e s Developed by Viscous Materials in the
Screw Extrusion Machine. ASME T r a n s . , vol. 73, Oct. 1951,
pp. 947-955.
14. Anon.: Theory of Extrusion, Ind. Eng. Chem., vol. 45, oo. 5,
May 1953, pp. 969-993.
15. Eccher, Silvio; and Valentinotti, Aldo: Experimental Determination
of Velocity Profiles in an Extruder Screw. Ind. Eng. Chem., vol.
50, no. 5, May 1958, pp. 829-836.
16. Griffith, R. M . : Fully Developed Flow in Screw Extruders. Ind.
Eng. Chem. Fundamentals, vol. 1, no. 3, Aug. 1962, pp. 180-187.
17. Squires, P . H.: Screw Extrusion - Flow Patterns and Recent Theo-
retical Developments. SPE T r a n s . , vol. 4, no. 1, Jan. 1964, pp.
7-16.
18. McGrew, J. M.; and McHugh, J. D.: Analysis and Test of the Screw
Seal in Laminar and Turbulent Operation. J. Basic Eng,, vol. 87,
no. 1, Mar. 1965, pp. 153-162.
19. Zotov, V. A . : Research on Helical Groove Seals. Russ. Eng. J . ,
vol, 10, Oct. 1959, pp. 3-7.
Asanuma, T . : Studies on the Sealing Action of Viscous Fluids.
Paper No. A3 presented at the F i r s t Mernational Conference on
Fluid Sealing. Cranfield, England, British Hydromechanics Re-
search Association, Harlow, Essex, England, Apr. 1961.
Lessley, R. L . ; and Hodgson, J. N . : Low-Leakage Dynamic Seal-
to~Space. Paper 65-GTP-14, ASME, Feb. 1965.
Whipple, R. T. P . : The Inclined Groove Bearing. Rep. AERE-T/
R-622 (rev.), Research Group, Atomic Energy Research Establish-
ment, United Kingdom Atomic Energy Authority, 1958.
Gruntz, Robert D . ; and Rackley, Ray A.: Snap 50/Spur Power Con-
version System-Objectives,, Current Status and Lunar Applications.
Paper No. 650321, SAE, May 1965.
Golubiev, A. L : Studies on Seals for Rotating Shafts of High-Pressure
Pumps. Wear, vol. 8, no. 4, July/Aug. 1965, pp. 270-288.
Holan, Karel: Sealing in Engineering. Proceedings of the Second
International Conference on Fluid Sealings, B. S. Nau, H. S.
Stephens, and D. E. Turnbull, e d s . , British Hydromechanics Re-
search Association, Harlow, Essex, England, 1964, pp. E5-73 -
E5-88.
Stair, William K.: Analysis of the Visco Seal. Rep. No. ME 65-587-2
University of Tennessee, Jan. 18, 1965.
Stair, William K.: Theoretical and Experimental Studies of Visco-
Type Shaft Seals. Rep. No. ME 66-587-5, University of Tennessee,
Apr. 28, 1966.
5.1.57

28. Lessley, R. L . : Snap-8 Seals-to-Space Development Test Program.


Vol. 1 - Visco Pump. Rep. No. 2808 (Topical) (NASA CR-54234),
Aerojet-General Corp., May 1964.
29. Kettleborough, C. F . : Turbulent and Inertia Flow in Slider Bearings,
ASLE T r a n s . , vol. 8, no. 3, July 1965, pp. 286-295.
30. Zuk, John; Ludwig, Lawrence P . ; and Johnson, Robert L . : Flow and
P r e s s u r e Field Analysis of Parallel Groove Geometry for an Incom-
pressible Fluid With Convective Inertia Effects. NASA TN D-3635,
1966.
31. Kuivinen, David E . : Determination of Oxygen in Liquid Alkali Metals
by the Mercury Amalgamation Method. Paper presented at the
Nineteenth Meeting, Chemical Rocket Propulsion Group (St. Paul,
Minn.), July-Aug., 1963.
32. Schlichting, Hermann (J. Kestin, T r a n s . ) : Boundary Layer Theory.
Fourth ed., McGraw-Hill Book Co., Inc., 1960.
33. Kulinski, E. S.; and Ostrach, S.: A Critical Evaluation of the Status
and Trends in High Speed Fluid Film Lubrication (to be published).
34. Bisson, E d m o n d E . ; and Anderson, William J . : Advanced Bearing
Technology. NASA SP-38, 1964.
35. Booy, M. L . : Influence of Oblique Channel Ends on Screw-Pump
Performance. J. Basic Eng., vol. 88, no. 1, Mar. 1966, pp.
121-13L
36. Langlois, W. E . : Slow Viscous Flow. Macmillan Co., 1964, pp.
105-107.
5.1.58

•I

TABLE I. - VISCOSITY
OF MINERAL OIL

Temperature Kinematic
.Op
' viscosity,
°C cs

100 37.8 71.2


130 54.4 32.4
210 98.9 8.4
r
K-3691

TABLE II ~ VI3C0SEAL GEOMETRIFS

Pr mary j r oove Sectr d a r / froov^


Groove Land R a t i o of Groove Radial Nur b e r Number Secondary Cross sect±of
width. width. groove ('epth. clearance. Dl n c ] i : ^ •>f h e L i x he] ix through land
to land •ir 1L, s art s starts angle.
Width, a.

W//////A /////A/////A
0 0
( 0 I 10 ir ) (O 1 0 (0 0

0 0 6 ^ en -, 0 Oo
V////////////////////A
0 5^
( jeo-"! (O ] 0

S'^C
f-^—-— ——•—i

V//////////////////////A

0 =(6 0 0 0 010
(0 10 I n ( 0 0^ in ) (0 Oi (0 00--

V////////////////////////i
O 01'-
(0 1 ) (0 0

se''0'"jar
(ircoves

wmm^^'^m^
0 4 0 0 0 0
( 0 ] iO i n ) I (O L30 (O a 0 4

Usec! a s b a b e v
TABLE m . - CRITICAL MODIFIED REYNOLDS NUMBER FOH ONSET OF TURBULENCE OF VAMOUS EXPERIMENTERS

Experimenter Reference Seal Diameter, Radial Ratio of Ridge Groove Helix C h a r a c t e r i s t i c C l e a r a n c e Modified
and fluid in. clearance, radial width, width, angle, length, Reynolds Reynolds
c, c l e a r a n c e to a, b, a, L, number, number.
in. r o t o r r a d i u s , in. in. deg in. R«c Re*
c/R

Stair 27 ai 1.2430 0.0042 0.0068 0.0934 0.1596 14.5 1.01 400 1.7
Water 2 1-. 2465 .00235 .0038 .1176 .0494 9.67 .98 (b) (b)
2B 1.2420 .0047 .0076 .1176 .0494 9.67 .98 (b) (b)
3 1.2461 .00265 .0043 .0828 .0842 9.67 .98 350 .95
3B 1.2408 .0053 .0085 .0828 .0842 9.67 .98 (b) (b)
4 1.2461 .00265 .0043 .0514 .1156 9.67 .98 200 .95
4B 1.2408 .0053 .0085 .0514 .1156 9.67 .98 400 2.2
5 1.2455 . 00295 .0047 .1585 .0607 5.81 1.83 600 .97
6 1.2455 .00295 .0047 .1070 .1083 5.81 1.83 200 .32
7 1.2455 .00295 .0047 .0609 .1600 5.81 1.83 160 .26

King 4 and 27 3C 2 0.0029 0.0029 0.1063 0.1063 3.86 3.2 550 0.50
Oil, IB 2 .0020 .0020 .1063 . 1063 3.86 3.2 400 .26
water, po- 2E 2 .0026 .0020 .1000 . 1000 7.26 1.54 650 1.1
tassium 5D 2 .0015 .0015 .1000 .1000 7.26 1.54 650 .63

McGrew and 18 S F 9 6 - 5 oil 1 0.00325 0.0065 (c) (c) 14.5 3.14 400 0.41
McHugh Oil

Lessley 28 Configura- 2 0.0032 0.0032 0.075 0.122 14.5 0.79 350 1.4
Oil tion B
Water Configura- 2 .0032 .0032 .083 .136 4 3.13 500 . 51
Mercurs^ tion C

^Eccentricity of rotor to housii^, c = 0 . 1 .


Break point not well defined.
^Four thread starts.
Slot explicitly found at breakpoint but falls on data line.
5.1.61

D
O
I
xl Liquid under
-Liquid-vapor ,-^.,Externally pressurized gas journal bearings
pressure -o
interface
Roller
bearings T-

Support shaft-
,, I \ •- Viscoseal rotor
^Viscoseal housing

Heat exchanger ^Externally pressurized


-Throttle valves
gas thrust bearing

Figure 1. - Viscoseal and hydraulic system.


5.1.62

^r-Groove depth, hg
/ 0.015 in.

05
CO
to
I
H

Housing twre,
2.001 in.

Section A-A
(a) Axial locations.

r A ^ ^ Pressure taps

m Circumferential locatfons.
Figure 2. - Pressure tap and thermcKOuple focations.
5.1.63

k Water out

• ^ Land width, a'

-Groovewidth, b'

Groove depth, h^
1
T
^ - Input
Support shaft Rotation shaft
CD
to
I

I Radial clearance, c

Helix angle, a, 141/2°


Water in

Figure 3. - Internal grooved housing viscoseal used with water.

Land width, a'


I To leak detector
n Groove width, b'
Liquid-gas interface-.

Argon gas
atmosphere
t

^ ^ probe / :

-Heated shroud
w -Resistance heater wires
^Resist -Helix angle, a

I Line to reservoir

Figure 4. - Grooved rotor viscoseal used to seal sodium.


5.1.64

Differentia
to transformer-|
to
I
Leak rate
monitoring
tank Air-operated
Argon gas valve
pressure

Hand valve
Float
Sodium level

'777777777TT77777777I272Z777777777/jy^/7y.

Figure 5. - Sodium supply system.


5.1.65

•• Plane orthogonal to
sealcenterline

Pressurized
cavity

Characterfstfc
length - ABC

Housing, m •* Rotor
nal grooves- rotation

Rotor
centerline

Plane

Pressure decreas-
ing over land

Rotor
Pressure increas rotation
ing over groove

Sealcenterline

Figure 6. - Isometric representation of viscoseal and pressure patterns.

MII-2066
Pressure tap

n Pressure decreasing
r i ci5ur c iMci ycjbmy uver groovy nvpr land
r 1^
-• Axial distance
f r n m r7i\i\^i

^-- - ' \ in.


• ^ .
X 0.33
^^ - ^ ...^- X \ '
• ^ ^^ \ •X.
\ - ^ Ji^ *•
.,^ ^ .--' ^
\ K ----
.53-,

^-^
. ^ ^ ^ - - \ \ \
.-^
X
^^^
.--- - ^ \ ^ 3- '
^ \ ^

(I) Rotational speed, 2000 rpm; average temperature of oil film, 46.2° C 1115° F); temperature
change along film, 0.8°C (1.5°F); cavity pressure, 30.4 Newtons per square centimeter
(44.1 Wsq in. absolute).

. 167r . 24ir . 32ir .40jr .4«7r .56ir


Circumferential distance, rad

(b) Rotational speed, 4000 rpm; average temperature of oil film, 138° F; temperature change along
film, 4° F; cavity pressure, 60.2 Ib/sq in. absolute.

Figure 7. - Pressure gradients on planes orthogonal to viscoseal axis centerline.


5.1.67

Housing,. --Land leading edge


Pressurized
cavity— ' —Land trailing edge
^Land
End of seal —Groove
0.19 in.
End
—effect H
lerigth
10
.-s- ' \ \ |
TO \
o \
\
\
\ \
.8
"~-E ndof hellce 1 \
groove s >
\ \
(a) Rotational speed, 1000 rpm; average temperature of oil film,
97° F; temperature change along oil film, 11° F; cavity pres-
sure, 54.1 Ib/sq in. absolute

1.1.
End 1 1
-effect * i 0.030 in. from land
len( th leading edge
.„„
0.030 in. from land
1.0
\ \ tr ailing edge

N
\
.9
\ \
•^ \
\\ \
.8 \
\
s
s
7 \
.1 .2 .3 .4 .5
Axial distance

(b) Rotational speed, 6000 rpm; average temperature of oil film,


132° F; temperature change along oil film, 19° F; cavity pres-
sure, 80.6 Ib/sq in. absolute.

Figures. - Axial pressure gradients near helical groove edges


with grooves in direct communication with pressurized
cavity.
5.1.68
I
4
r^"^^m^

High-
pressure I
end | r ^ Liquid-gas
interface

t Rotation

^Rotor

j ' " ~ Housing


with
internal
|c-65-687| '"^''
_Continuous fluid film, ,^Scavenging
(no gas ingestion) or non-
wetted length •\
(a) Modified Reynolds number, 1.75 (2000 rpm).

High-
pressure
end

ging or •
nonwetted
-« Ingestion direction ''"^tti

(b) Modified Reynolds number, 2.58(3000 rpm).

Figure 9. - Grooved housing viscoseal operation when sealing water.

MTI-2071
I 5,1.69

I*

High-
pressure
end

-Ingestion
direction
(c) Modified Reynolds number, 3 50(4000 rpm)

High-
pressure
end

I Mixture of gas ^ | ^ Discontinuous ^ | Scavenging or


r*"bubbles and water*T*~fluid film -*T*nonwetted -
I I I length
-Ingestion
direction
(d) Modified Reynolds number 4 42 (5000 rpm)

Figure 9 Continued
> •

MTI-2073
5.1.70 I
•I

High-
pressure
end

wetted
-Ingestion direction length
(e) Modified Reynolds number, 5 25 (6000 rpm)

High-
pressure
end

Continuous fluid Gas bubbles Scavenging or

h film (no gas


ingestion)
•and discon •
tinuous film
-nonwetted
length

(f) Modified Reynolds number, 7 00(8000 rpm)

Figure 9 Concluded

••
MTI-2072
H
CO
to
I

2 3 4 5 6 7 8
Modified Reynolds number

Figure 10. - Comparison of observed gas ingestion


when sealing water for grooved housing and
grooved rotor viscoseals.

Gas bubbles Primary Gas bubbles


Primary
groove-. groove Housin

Plane orthogonal to viscoseal axis Plane orthogonal to viscoseal axis

(a) Grooved housing. Bubble motion pri- (W Grooved rotor. Bubble motion has axial
marily cirtumferential. component due to helical grooves.

Figure 11. - Gas ingestion mechanisms of grooved housing compared with grooved rotor viscoseal.
72

^^^^K^'

High-
pressure
eno
^

itiJ^l
1 ^

-«—-Gas ingestion direction and -No leakage at


increasing liquid-gas density low-pressure
end
Land trailing edge
Land leading edge

End of scavenging
i 3h

£ 2h Q. ^,,,»~P essure gradient in-


^y' ( eases with Increasing
' ' ^ quid-gas density

Axial distance, in.

Figure 12 - Pressure as function of axial length wi th gas ingestion for


grooved rotor Sealed fluid, w/ater

MII-2067
5.1.73

Modified Reynolds number

Figure 13. - Gas ingestion rate when sealing liquid sodium at 400° F.
i_ To leak detector

pSealed liquid
Direction of
gas blowout

"^Liquid-gas interface
, '-Gas bubbles ingested towards high-pressure end
t \ . "-Gas bubble increasing in size
^—To reservoir
(see fig. 5)
(a) Mechanism involved in gas ingestion into pressurized cavity.

Gas ingestion Gas blowout


indication (in- when bubbles
crease in leak size equals >- Zero gas
detector level)-7 rotor diameter-r-. ' ingestion;
/ I zero leak
/
line

(b) Schematic liquid level oscillograph trace for repeated cycles of gas ingestion, subsequent
gas blowout, and no liquid leakage at blowout.

-Gas blowout
Gas ingestion-, Zero gas ingestion;
I

Leakage at blowout

Time

(c) Schematic liquid level oscillograph trace for repeated cycles of gas ingestion, subsequent
gas blowout, and liquid leakage at blowout.

Figure 14. - Gas ingestion into closed cavity.


5.1.75

^
^
•r^ ^ y -
._^¥ ^V
ZZ_B: /
y /
p <- '
^Z23
^I
U
1 1 1 M M 1
—o— Smooth-bore housing,
E
to (geometry 1)
to D- Smooth rotor, five grooves
in housing (geometry 5)
.1 i 1 1
103 10^ 105
Reynolds number based on clearance

Figure 15. - Comparison of sealing parameters for grooved rotor and housing
operating in sodium. Sodium temperature, 165° to 335° C (329° to 635° F).

P0)
dl
r^sr
~'a"
«
r/" '
CH Z J
V
r-ija
.•^L / X,
.B o t
i
r ,{/
c/^

I 1
—c) S moot i-b<)re hous ng, five groov eson •otor (ge 3m ;try 1)
— c ] — - S moot i-b<)re hous ng, five groov esand secondary and
gro ove >o 1 rotor (geometry 2)
,1 1 1
(a) Five helix starts.
1— [ —

J / ^'
55
y
^f'X}
j ^
M> r 7Sr^
r .
/ o (Y X)
, 1 ^ ^

C
'^ cf'
& o
1 1 M ' 1 1 1 1 1
— O — Smooth -bor e hous ng, 10 grooves on rotor (geometry . *
—n—Smooth -bor e hous ng, 10 grooves and secondary land
groo ves on roto r (geometry 4)
1 1 L_L
M 1 , 1 1 1 11
103 10^ 105
Reynolds number based on clearance

(b) Ten helix starts.

Figure 16. - Comparison of sealing parameter obtained with helically grooved


rotor with and without secondary land grooves operating in sodium. So-
dium temperature, 165° to 335° C (329° to 635° F).
5.1.76

CD
^0

_ .jmferential
I unwrapped cylinder

(a) Plan view.

(b) Cross-sectional view along z-axis illustrating relative


motion of smooth flat plate with respect to fixed parallel
groove-ridge plate.

Figure 17. - Mathematical model of parallel groove-ndge geometry of infinite horizontal extent
•»—Scavenging or
nonwetted region
Cavity
pressure

Average pres-
sure ridge
leading edge
Averse pres-
Pressure sure ridge
trailing edge

,,-Ambient pressure

Axial distance

High pres-
sure end

C-66-2693
Figure 18. - Region of analysis validity (homogeneous fluid) of a grooved housing viscoseal, Reynolds
number based on clearance or film thici<ness, 1700; modified or reduced Reynolds number, 19;
400-microinch shaft rotational movement during photographic exposure; shaft speed, a)Oo'rpm;
sealed fluid water
High pres-
sure end

mw-'^'

_Discontinuous fluid film_


(mixed air and liquid)

Figure 19. - Viscoseal with grooved housing operating in a regime of water-air mix-
ture (gas ingestion). Analyses does not apply for this discontinuous fluid film
case; 250-microinch shaft rotational movement during photographic exposure;
shaft speed, 5000 rpm, sealed fluid, water

U sin a

c , (3) 1 (3)| y

^ ^

Figure 20 - Cross-sectional view along z-axis


illustrating arbitrary groove-ndge pair shape
5.1.79

2irR

H
cn
to
to

Circumferential

Figure 21 - -Resolution of across and along groove-ndge pressure gradients into


axial pressure gradient "Unwrapped" cylinder is shown

Leading^
edge-/" y;/^f ^

CD-8606
Figure 22 Pressure distribution over parallel groove geometry (Qualitative
graphical representation from experimental data )
5.1.80

£ ^ ^ '^''fe^ leading edge;^,


^Reference
/ pressure
en
CD
I

nLcosa
Coordinate along ridge-groove, x

(a) Along ridges or grooves (no end effects are shown).

Reference
pressure-

nL sin a -
f^a -^
Ridge'"
Groove
Coordinate across ridge-groove, z

(b) Across ridges and grooves (no restric-


tion that profile must be linear).

Figure 23. - Pressure profiles (qualitative graphical representation from experimental data)
E-3691

Seal Width Radial Helix Eccentricity o


ratio, clearance, angle, ratio
b/a c, a. cq.
in. deg
• 5 0.38 0.00295 5.81 0.3 -^
A 6 1.0 .00295 5.81 .3
O 7 2.6 .00295 5.81 .3 5iA-^^
,€-^
-^fei
.H3°
: ^ ^A £
-^i ^
°.c^
.^ ^ : ^ ^^ ^
- - . ^ r ^ ^ -
- a^ ^
^:ni:;:r•-Drrrf>
-'•-Breakpoints

- -o- -TT
"Breakpoint:

-c
-&- rg3:
i^tt :»is^
fe
60 80 m 160 200 400 600 80) low 2(M0 40)0
Clearance Reynolds number

Figure 24, - Experimental sealing parameter for seals 5, 6, and 7 from reference 27.
1 1
Significant /
convective /
inertia y
H>-Actual region effects /
of creeping
flow validity yf Generally accepted by
y// current researchers

/ inertia effects begin


1 'S gradually
A
^Breakpoint
(onset of turbulence)

Claimed creeping Turbulent flow regime


flow regime (claimed by present
day researchers)
Re ~ P
—s* ~
~0 01 Critical
Clearance Reynolds number Re,, (log scale)
Figure 25 Empirical sealing coefficient as function of
clearance Reynolds number

^-Representation of the helical path


' of a typical ridge (land)

Repre-
sentation
of top view

.^^

Schematic
representation
of end view

Figure 26 Illustration that a point on the viscoseal surface can be


reached from the origin axis along both the x constant and z
constant surface coordinate paths (for example paths from o to a
5.1.83

to where Pj > PQ
Po ^
to
I Section A-A of groove-ridge pattern
U
gg????S^S
^^m^^^^^^
x ^ ^ ^ ^mm^^^M^^^^^^^.^ ^
Figure 27 /Vlodel used m creeping flow solution

(High pressure)

Ux-0.96IJ,
Uj-0.26 0

(Low pressure)

Figure 28 - Resolution of plate velocity lor drag force)


into components along and across groove or ridge

TzMzimzmzzzzzm^^^^^^zzzzzzmzzzzzmzim.
m>.
VWZ^^^^W7Z^7777777'^7Z^77?7Zy7Z^^^
1000 \ or 1000 c

Figure 29 - Example where the creeping flow mathematical model can be considered a valid
physical model Modified Reynolds number (Re ) much less than 1 since clearance c is
much less than characteristic length L

NASA CLEVLLAND OHIO E 3691


5,2, I

THE EFFECT OF TURBULENCE ON VISCO SEAL PERFORMANCE


by
WiI I iam K. Sta i r
Professor of Mechanical Engineering
University of Tennessee

ABSTRACT
Experimental study of the visco seal reveals that the laminar seal performance
is in excellent agreement with the analysis of Boon and TaI. In the turbulent
region, however, the sealing performance exceeds the theoretical laminar pre-
diction and the optimum laminar seal geometry is not optimum for turbulent
operation. A theoretical equation has been obtained for the sealing coefficient
of a visco seal which is suitable for both laminar and turbulent flow. Experimental
results from 13 different seal geometries confirm the utility of the analysis.

Air ingestion has been observed during turbulent operation of each experimental
seal. This phenomenon, however is not fully understood, may become a major
problem in the application of the visco seal.

I. INTRODUCTI ON
The viscoseal, viscosity pump, screw seal, spiral groove seal or bearing are various
names given the device whose working principle is based on the pressure generation
in a viscous fluid, enclosed in a narrow annulus or slit, by means of grooves on a
rotating shaft or plate. The viscoseal, shown basically in Figure I, is an old but
little used device which functions without sliding contact and, unlike other
clearance-type seals, holds promise of zero leakage. The lack of contact, long
reliable life, and the possibility of zero leakage has recently created a renewed
interest in the viscoseal as a sealing element for use in the critical applications
associated with the nuclear energy and space programs (I).

I I. BACKGROUND
The performance of the viscoseal may be expressed in terms of a sealing coefficient
defined as:
A = ^
c Z^

A number of theoretical analyses have been presented for laminar operation of


the viscoseal with homogeneous fluids in which the theoretical laminar sealing
coefficient is presented as a function of groove geometry (2). The grooves
shown on the shaft in Figure I, may be located on the shaft or housing insofar
as laminar theory is concerned. These analyses lead to

Numbers in parenthesis designate references at end of paper.


expressions for the sealing coefficient which a r e , or can be modified to
b e , in the form:

A = ^^^^ = y"(seal geometry). (2)


C AP

Sealing coefficients determined experimentally for a number of sealing


geometries may be characterized as follows: ( 2 , 3 , 4 , 5 , 6 )
1. At low Reynolds numbers the sealing coefficient is constant (inde-
pendent of Reynolds number) and is a function of the groove geometry.
The experimental laminar sealing coefficients are in excellent agreement
with the equation,

^ t T ( 1 - ?- ) ( s ^ - i )
which is based on the work of Boon and Tal (7) . Figure 2 presents a
comparison of several laminar analyses with data for t e s t seal No. 1.
2. At high Reynolds numbers the sealing coefficient improves (/V
decreases) and is a function of Reynolds number as well as seal geometry.
3. There is a smooth systematic transition from laminar to turbulent
operation and the Reynolds number at which the transition begins is a
function of seal geometry. As T is increased in a seal t e s t s e r i e s ,
having the same a and B, the onset of turbulence begins at lower
Reynolds numbers. These observations suggest that whatever the
theoretical expression for the turbulent A might b e , it should become
equivalent to equation (3) at low values of Reynolds number.
The a n a l y s e s of the turbulent v i s c o s e a l are very limited and are
largely empirical. McGrew and McHugh (5) made use of the Prandtl
mixing length concept and deduced the form of the turbulent sealing
coefficient a s :

Ao=«=l*<=2Re^)-l (4)
Since the experimental constants were determined for a single seal only,
equation 4 is of limited utility in design.
TURBULENT SEALING COEFFICIENT
Making use of the same physical model employed by Boon and
Tal (7) for the laminar a n a l y s i s , a theoretical sealing coefficient for the
turbulent v i s c o s e a l has been derived (2). The procedure utilized in the
turbulent a n a l y s i s was to modify the equations of motion and continuity
used in the laminar a n a l y s i s by replacing instantaneous velocity
components and pressure with mean components and turbulent fluctua-
ting components in order to represent turbulent flow. From these basic
equations the Reynolds equations for turbulent flow were formed.
Following a method similar to that employed by Pai, and representing
the velocity profiles for Poiseuille and Couette flow as power s e r i e s ,
the equations for the Poiseuille and Couette velocity components were
obtained. The flow components in the seal were the same as those
identified in the laminar a n a l y s i s (9). Setting the sum of the flow
components to zero, the turbulent sealing coefficient was obtained a s :

h^h r 11
3
A^ = h + K,.
5 I.
(5)
L 4
where

I^= ( 1 - ^)t, (6)

3 9
(7)
ig - p r t ,

^3 = PV [ y + P ^ I - t )] (8)

and

y + ^ (1 - y ) + g(p - 1)
^4 = ^ [l- ^ + ¥ (9)
I + 6^1 - t )
The terms I , I , I , and I . are functions of the seal geometric parameters ^
X Zi sj T:

a , P, and y and are independent of Reynolds number. K and K ,


however, are functions of seal geometry and Reynolds number and are given
by:
1 - F{
'I
1 - +
^4""2F 10. 5F ~ 7.5
I S 3.92F - 1.4F^
5
(10)
and
1 - F.
1 - . (11)
S " 2F^ 10. 5F^ '7.5
3.92F, - 1.4F, 1 -^
K., which is a function of F ^ relates to flow in the direction parallel
4 5
to the grooves, while K^., a function of F ^^ , is related to flow in the
direction normal to the g r o o v e s . The factors F g and F,,. are expressed
as the ratio of the wall shearing s t r e s s in turbulent flow to the wall
shearing s t r e s s for laminar flow with the same maximum channel velocity,
Thus,
'(t/T,)^ =( / / / , ) (12)
I
and

n = ( V ^*'n =< ^°/^/'


o" ~ ^ • n. • W - / - K ,
(13)

The formulation of equation 5 considers that the onset of turbulence in the


grooves and over the lands does not occur at the same Reynolds number.
The Taylor criterion for the onset of turbulence in concentric cylinders
with the inner cylinder rotating is

Re 41.1 V "^ (14)

In this a n a l y s i s , an average critical Reynolds number for the visco seal


was defined a s :
1/2
D/2
Re .5 41.1 (15)
crit (1 - / )C -F yf3c
At this juncture the coefficients K and K may be evaluated in terms of
the experimental data which have been obtained. However, it was observed
early in this program that the plots of the friction parameter versus
Reynolds number for the visco s e a l were similar in general appearance
to plots of the friction factor versus Reynolds number for ordinary pipe
flow. In an effort to utilize readily a v a i l a b l e , through indirect, data
to predict the turbulent visco seal performance, K. and K^ were
estimated by the use of Figure 3 . In this figure the commonly used
friction factor for laminar flow in pipes is expressed by the equation:

P
In the turbulent range the coefficient of resistance for flow normal to
the grooves was expected to be different from the coefficient for flow
parallel to the grooves. In Figure 3 two lines were selected to
represent estimates of the r e s i s t a n c e coefficients for these two flow
directions. The lower l i n e , defined by the equation

f=0.326/Re ^-^^^, (17)


P
w a s selected to represent the friction factor for flow parallel to the
g r o o v e s . This curve was approximated by extending the conventional
"smooth pipe" curve to the intersection with equation 16. The upper
curve, defined by the equation
f=0.646/Re 0-246^ ^gj
P
was selected to represent the friction factor for flow normal to the
grooves. This curve was estimated by extending the curve for a rough-
n e s s ratio of approximately 0 . 0 3 , which corresponds to that for corru-
gated pipe, to its intersection with the laminar equation 16. This
procedure, of c o u r s e , represents the transition region as straight l i n e s .
Assuming the critical Reynolds number for pipe flow to be 2 , 0 0 0 ,
the term F t is evaluated, according to equation (12), by determining
the ratio of the friction factor for turbulent flow to the friction factor
for laminar flow, both at the pipe Reynolds number of:
Re = (2000 cos a) Re /Re . . (19)
p c crit.
Re refers to the seal Reynolds number based on the clearance and
c
Re . is defined by equation (15). For evaluation of F v the lower
crit. 5
turbulent line on Figure 3 is u s e d . F^^ is determined in a similar
manner except at a pipe Reynolds number of:
Re = (2000 sin a) Re /Re ,^ , (20)
p c crit.
and using the upper turbulent line on Figure 3 . The coefficients K. and
K can now be evaluated by equation (10) and (11) r e s p e c t i v e l y .
When the seal is operating in the laminar range K. and K are
both unity and equation (5) becomes equal to equation (3) . As the
Reynolds number i n c r e a s e s F t i n c r e a s e s , causing K. to d e c r e a s e ,
and the sealing coefficient begins to d e c r e a s e . With further i n c r e a s e s
in Reynolds number F^ a l s o begins to increase thus causing K,. to
decrease which produces a further decrease in the sealing coefficient.
GENERALIZING THE THEORETICAL SEALING COEFFICIENT
Equation (5) was evaluated for a series of s e a l s having helix
angles of 5 . 8 1 ° , 9.67° , 1 4 . 5 ° , and 20. 15° for Reynolds numbers of
100, 500, 1000, and 5000. In this general c a s e the seal diameter and
clearance were fixed at 1.250" and 0 . 0 0 3 " respectively. Typical results
of t h e s e calculations are plotted as Figures 4 through 8. Several obser-
vations may be made regarding t h e s e figures:
1. The optimum screw geometry in the laminar range is not the
optimum geometry for turbulent operation. As shown in Figure 4 , a seal
having an a of 14.5° has a minimum sealing coefficient for laminar
operation when / = 0. 5 and 3 = 4 . However, when operation becomes
turbulent the minimum value of the sealing coefficient is found at higher
values of ^ and P. At higher a values the same observation can be made.
However, the shift in the optimum P becomes smaller.
2. The p producing the minimum sealing coefficient in the
turbulent range i n c r e a s e s with increasing Reynolds number. This effect
is l e s s significant for large a than for small a .
3. In both laminar and turbulent flow, a given change in P
produces a greater change in A m with screws having large a than
with small a . Thus, from the standpoint of pressure stability the lower
values of a are to be preferred.
4. In laminar operation the optimum t for any screw is 0 . 5 .
However, in turbulent flow the sealing coefficient is improved slightly
as y is increased from 0 . 5 to 0 . 7 .
5. For s e a l s which must operate in both the laminar and tur-
bulent regions, the smaller helix angles will give the best values of the
sealing coefficient for both regions.
6. The transition from laminar to turbulent operation is a
function of y . As J i n c r e a s e s the onset of turbulence occurs at
lower Reynolds numbers.
7. As the visco seal operation begins to be turbulent the effect
of ^ on Arn ^t the larger P values is rather pronounced. However,
the degree of turbulence i n c r e a s e s this effect becomes l e s s pronounced.
8. At a Reynolds number of 100, A ^ is the same as the one
computed in equation (3), which is independent of Reynolds number.
EXPERIMENTAL FACILITIES AND TEST PROCEDURE
The visco seal t e s t section and drive is shown in Figure 9, and
a schematic diagram of the t e s t section is shown in Figure 10. The
sealant fluid, which was distilled water in all t e s t s reported, was
introduced under regulated pressure through the sealant inlet. Pressure
t a p s and thermocouples are located along the axis of the t e s t s l e e v e ,
the dimensional location of which are shown in Figure 11. Inductance
probes located in two planes and 90° apart are used to determine the
eccentricity between the t e s t sleeve and spindle. A torque arm con-
nected to a strain gage bridge is used to measure the frictional torque.
Since the torque measured under some operational conditions is very
low, the t e s t sleeve is mounted on eight hydrostatic bearing pads to
minimize static friction. The dimensions of the t e s t spindles are pre-
sented in Table I . Test spindles 1, 2, 3, 4, 2B, 3B, and 4B, shown in
Figure 12, were threaded the total seal length. Spindles 5, 6, and 7,
shown in Figure 12, were constructed with end dams in order to study
the end effects of the seal and the phenomenon of air ingestion.
During a t e s t run the data recorded included spindle speed,
torque, eccentricity, pressure distribution, and temperature distribution.
Using the information in Table I, the recorded experimental d a t a , and the
physical properties of the t e s t fluid, the experimental sealing coefficient
(hereafter referred to a s A p ) ^^'^ the friction parameter were calculated.
Lambda and friction parameter were plotted versus Reynolds number
based on the c l e a r a n c e .
The quantity AP/L in equation (2) for A p ^^Y be interpreted
in two w a y s . In a practical application this value should be (P -
P , ) / L . However, in this study the effect of screw geometry
atmosphere
on the sealing coefficient was a factor of major concern. Therefore, in
order to minimize the end effects, the A P / L was interpreted as dP/dL and
evaluated a s shown in Figure 11. The viscosity of the t e s t fluid was
evaluated at an average temperature determined from the distribution a s
shown in Figure 1 1 . The temperature variation in the seal was quite small
being on the order of 2 degrees F,
EXPERIMENTAL RESULTS
Sealing Coefficient:
The geometry of spindle number 1 was made similar to the one
employed by McGrew and McHugh for the purpose of comparison of
experimental r e s u l t s . The data compared favorably in the turbulent
range, a s shown in Figure 13, to the experimental curve of McGrew and
McHugh. In the laminar range, however, the value of McGrew and
McHugh's experimental curve is higher than the present d a t a .
A summary of t e s t results is presented in Table II. Using the
theory presented in section III, the theoretical sealing coefficient is
compared with the experimental values in Figures 13 through 18.
For all spindles t e s t e d A -p remained e s s e n t i a l l y constant
during laminar operation. This result is in agreement with equation (3)
which indicates that y \ _ / f(Re ) in the laminar region. The average
ratio of A ^ to A q, for laminar operation is 0 . 9 6 8 . During the
turbulent operation of each of the s e a l s 2, 3, 4 , 2B, 3B, and 4B, the
ratio of A-n to A „ remained essentially constant. However,
because of the variation of slopes in the turbulent region for t e s t series
5 , 6 , and 7, the ratio of A p to A.^ was not constant.
With the onset of turbulence A p begins to decrease with
increasing Reynolds numbers. The Reynolds number at which transition
from laminar to turbulent operation occurs is a function of 3 . In t e s t
series 2 , 3 , and 4 , the three t e s t spindles have essentially the same
a, p , c , and D with variations in T only. Spindle 2, having the
lowest / , resulted in the longest laminar operational region, while
spindle 4 , with the largest T , resulted in the shortest operational
laminar region, a s may be observed in Figure 14. Thus, as F increases
the transitional Reynolds number d e c r e a s e s . The decrease in the t r a n s i -
tional Reynolds number, defined in equation (15), is to be expected since
the average clearance in the seal is increasing. This direct relationship
between the onset of turbulence and "f is in agreement with the theo-
retical projection in section III.
The relationship between a 4 A p was demonstrated in t e s t series
5, 6, and 7. The geometric parameters of spindles 5, 6, and 7 are
e s s e n t i a l l y the same as for spindles 2 , 3 , and 4 with the exception of
a which was decreased from 9.67° to 5 . 8 1 ° . In t e s t s of s e a l s having
a ' s of 9.67 and 1 4 . 5 ° , the slope din A „/cilnRe , in the turbulent zone
h C

was approximately - 0 . 7 0 , which agrees with the theoretical prediction


of-0.709. However, for an a of 5.81° the value of din A VdlnRe
h C

was found to be approximately - 0 . 4 3 for the range of experimental data


recorded. The relationship between A^p snd a is demonstrated in
Figure 8. In this figure it is observed that Arp. for a equal to 5 . 8 1 ° ,
remains laminar to a Reynolds number of 300, while for a equal to 20.15° ,
5.2,10

J\ begins to decrease at a Reynolds number of approximately 100.


Therefore, a s a i n c r e a s e s the transition point from laminar to turbulent
operation for J\ decreases. However, it is noted from Figure 8
t h a t , for either large or small a ' s , the predicted slope, din Ap/cJlnRe ,
for Reynolds numbers greater than 1000, is the same. In this regard
A p and A T ^ ° '^^t a g r e e .
The effect of P on J\ may be observed in Figures 16, 17, and
18. In series 2B, 3B, and 4B oc and y remain unchanged from series 2,
3 , and 4 , while P was decreased from approximately 7 to 3 . A p for
2B, 3B, and 4B i s , on the a v e r a g e , 1.29 times the values for spindles 2,
3 , and 4 in both laminar and turb ulent operation. This ratio compares
favorably with theoretical predictions which indicate that the ratio should
be 1.28. In t e s t series 2B there is a sudden large increase in A p at
a Reynolds number of approximately 1600. This discontinuity in A p is
believed to be the result of air ingestion which produces a decrease in
the value of P. A similar observation was made for spindles 3B and 4B,
but the perturbation was much lower in magnitude.
Friction Parameter and Power Loss:
The power loss in the visco seal during laminar operation may be
described in terms of a dissipation function (8) defined by the equation:

P p-^ (1 + t^) + t^ X (1 - ^ ) (P^ - 1)^

<i> = ^ 2 ' ^22)


% UD L U
where q is the power l o s s . The dissipation function depends upon a , p ,
and Y and is independent of rotational s p e e d . The friction parameter
may be defined in terms of the dissipation function a s :

F.P.= ^ . (23)
c
5.2.11

Thus, F . P . is a function of the seal parameters and speed. The laminar


experimental data for the friction parameter are in agreement with the
equation (23) a s shown in Figures 19, 20, and 2 1 . The experimental
data in the turbulent range follows the trend noted by Smith and
Fuller (lO) whose data for the friction parameter of a plain journal
bearing operating under laminar and turbulent conditions can be expressed
as:
F.P.^ = 4jr/Re , (24)
/ c
and
F.P. = 0.156 Tc/Re ° ' ^ ^ . (25)
o ., c
As the visco s e a l more nearly approaches a journal bearing, T becoming
smaller, the experimental data for the friction parameter approaches the
laminar experimental curve presented by Smith and Fuller. In the turbulent
region, however, the experimental data for the seal are very slightly
higher than that of Smith and Fuller. Thus, in laminar operation the
friction parameter is described by equation (23) and during turbulent
operation Smith and Fuller's turbulent journal bearing results can be
utilized to estimate the friction parameter.
Effect of Eccentricity on Sealing Coefficient:
McGrew and McHugh ( 5 ) predicted that eccentric operation of
the visco s e a l could cause as much a s a fifty percent increase in the
sealing coefficient obtained under concentric conditions. For laminar
operation in the absence of gas ingestion it was found that A p was
e s s e n t i a l l y independent of the s e a l eccentricity in the range of eccen-
tricity ratios up to 0 . 6 . In the turbulent range A p was observed to
decrease slightly with increasing eccentricity. This effect is shown in
Figure 13,
Gas Ingestion:
Extensive observations of a phenomenon which has been termed
"gas ingestion" have been made by Ludwig and others (4) . Other
•I
investigators have a l s o observed seal leaks or instabilities which may be
related to gas ingestion. McGrew and McHugh (5) observed small
leaks past the s e a l interface which were called "seal b r e a k s " . These
leaks occurred at different Reynolds numbers for the various t e s t fluids
u s e d , and the fluid surface tension was believed to be a significant
property controlling the s e a l break point. King (ll) referred to a
similar observation a s a "secondary l e a k , " and indicated that the leak
point was a function of speed and seal clearance and independent of seal
p r e s s u r e . An actual leak was not encountered in any of the t e s t s repre-
sented by Figures 13 through 18. However, t h e s e t e s t s were limited to
Reynolds numbers below which the pressure gradient remained stable for
the 15 to 20-minute period required to complete a single t e s t point.
During t h e s e apparently stable runs gas ingestion was usually observed
when in the turbulent z o n e .
While the exact c a u s e of gas ingestion is not known, the experi-
mental observations made in the study support the hypothesis first stated
by Ludwig ( 4) , that the portion of the grooved shaft which is exposed
to gas tends to pump the gas through the liquid-gas interface. The gas
and liquid become mixed, thus having lower density, and the effective
length of the s e a l i n c r e a s e s slightly and the pressure gradient at the low
pressure end of the seal is reduced. The centrifugal force in the fluid
element c a u s e s the gas bubbles to migrate to the root of the groove.
The changes in s e a l performance caused by gas ingestion appear
to depend upon the seal geometry. For example, seal 2B experienced a
sudden rise in the sealing coefficient as shown in Figure 1 1 . This rise
was quite reproducible since a number of excursions were made through
this region. Similar discontinuities were experienced with other t e s t
s e a l s but to a smaller d e g r e e . Test seal 2B had a P = 2.96 and a Y of
0.296, Note in Figure 6 that the slope of the A versus P curve is very
steep at low values of P. Thus, the accumulation of a small amount of flL
gas at the root of the grooves could effectively reduce P and thus cause
5.2.13

A to suddenly r i s e . Seals 3B and 4B had P which were almost identical


to that for 2B, yet t h e s e s e a l s exhibited only slight increases in A
during turbulent operation with gas ingestion. Seals 3B and 4B, however,
had r values of 0.504 and 0.692 respectively suggesting that the wider
grooves may be l e s s susceptible to gas accumulation at the thread root.
The question naturally a r i s e s as to whether the threaded shaft
visco s e a l can operate for extended periods in the turbulent region when
gas ingestion is occurring. Beginning with t e s t seal 2B, a series of
long term t e s t s was initiated to gain insight into the problem of turbulent
operation with gas ingestion. After a sufficient number of normal t e s t
runs are made to construct a performance curve, the seal is then subjected
to one or more extended runs during which the effect of gas ingestion is
observed. The s e a l t e s t facility is not presently instrumented to make
a quantitative measurement of gas ingestion. Rather, the pressure
gradient in the s e a l is monitored to observe gradual changes which occur
due to gas ingestion and to note the occurrence of pressure p u l s a t i o n s .
Also changes in the sealing coefficient at the start and end of the
extended t e s t run are noted, A long term t e s t is continued at the same
operation conditions for a period of one to two hours or until severe
pressure fluctuations are encountered. The long term t e s t is run under
two system conditions. First, the seal is operated with the pressure-
regulated sealant supply system connected to the s e a l supply inlet.
This condition produces a regulated pressure cavity having a volume of
239 in^. Then the seal is bled and operated with the supply system
valved off which produces a trapped volume of 4.9 in^ and the pressure
in the trapped volume is determined by the s e a l . The location of typical
t e s t points for long term t e s t s are shown in Figures 16, 17, and 18.
Test s e a l s 5, 6, and 7 were machined with small clearance dams
on each end of the grooved section of the shaft a s shown in Figure 12,
"^P These dams were expected to have two effects ( 12) : The inner
restriction is expected to reduce the end effect and the outer restriction
5.2.14

is used to reduce gas ingestion. Following the t e s t s with s e a l s 5, 6, and


7 the clearance dams were removed and the t e s t series repeated a s 5D,
6D, and 7D. The r e s u l t s of the latter t e s t series indicate that the end
dams do not adversely affect the sealing coefficient, the end effect is
reduced but not eliminated, and the severity of gas ingestion i s reduced.
The study of gas ingestion c o n t i n u e s . Though not complete, the
observations which have been made may be summarized as follows:
1. Gas ingestion under laminar conditions is so low a s to be undetect-
able in the present t e s t facility in either normal or long term t e s t s .
The pressure gradient in the seal remains e s s e n t i a l l y constant and
the sealing coefficient shows no significant change during extended
runs. Operation of the seal is not affected by changing from a large
regulated pressure supply to a small non-regulated trapped volume
at the high pressure end of the s e a l .
2. As speed of operation is increased a point is reached at which gas
ingestion can be observed. The Reynolds number at which gas
ingestion is first encountered is not the same for each s e a l . The
beginning and intensity of gas ingestion appears to be a function
of s e a l geometry. After gas ingestion has started the operation of
the seal may be affected by the volume and pressure regulation of
the sealed c a v i t y . Increasing the s e a l eccentricity ratio i n c r e a s e s
the intensity of gas ingestion.
3. The seal may operate in a stable manner with gas ingestion in the
turbulent region. The maximum Reynolds number for stable operation
is different for each s e a l . At low turbulent Reynolds numbers the
seal performance changes during a long term test and may reach a
stable operating point at a slightly higher sealing coefficient than
is obtained in a normal short term run. See Figures 16 and 17.
Under t h e s e conditions the pressure profile changes slightly and
the effective seal length i n c r e a s e s . ^&.
4. As the Reynolds number i n c r e a s e s a point is reached at which the
Seal pressure gradient becomes u n s t a b l e . When this occurs the seal
performance may show one of perhaps several c h a r a c t e r i s t i c s . The
following have been identified:
(a) The seal pressure profile experiences small fluctuations
followed by a substantial drop in the sealed head after which
a new stable operating curve is reached which permits operation
at even higher Reynolds numbers. This is the characteristic
exhibited by seal 2B (See Figure 16).
(b) The seal pressure profile experiences small low frequency
fluctuations accompanied by an increased effective seal length.
The s e a l will then continue in a steady low frequency pressure
fluctuation. If the sealed head is increased the effective
s e a l length required exceeds the threaded length and a small
more or l e s s steady leak occurs and continues until the sealant
supply is e x h a u s t e d . If both speed and the sealed pressure are
increased the small steady pressure fluctuations will increase
in amplitude and frequency and the leak becomes pulsating.
(c) The seal pressure profile exhibits a high frequency fluctuation
with an increasing amplitude and a pulsating or spurting type
of leak o c c u r s . This condition is more characteristic when
operating with a small trapped sealed cavity then when using
the large regulated cavity.
Other Experimental Results:
In Figure 22 the experimental data taken from a series of t e s t s
conducted by King (ll) , along with fy computed by equation (5)
for the thread geometry used by King, are shown. The predicted s e a l -
ing coefficient is in good agreement with the 5D and 2E series of data
in the turbulent range. In the laminar range the predicted value for
the sealing coefficient is smaller than King's extrapolated data c u r v e s .
It is difficult to draw conclusions about the laminar zone since King
obtained limited data in this region. In series 3C and IB, the slopes
of King's experimental sealing coefficients are - 0 . 5 2 7 and - 0 . 4 6 6 ,
agreeing with the slopes of A p in t e s t series 5 , 6 , and 7. Data curves
5D and 2E have slopes in the turbulent region of - 0 . 7 0 6 and - 0 . 6 8 6 ,
which are similar to the values obtained in t e s t with spindles 1 through
4B. From the experimental evidence reported in this work, along with
data from King, the indication is that for small helix a n g l e s , 3° to 6 ° ,
the slope of the experimental sealing coefficient in the turbulent zone
is approximately - 0 , 4 6 . For helix angles of 7° to 14° corresponding
to the range of optimum a recommendations of various i n v e s t i g a t o r s , the
slope of the experimental sealing coefficient is approximately - 0 . 7 0 .
CONCLUSIONS AND RECOMMENDATIONS
Based on the analytical laminar a n a l y s i s of the visco s e a l
developed by Boon and Tal, equations were derived to predict the operation
of the visco s e a l in the laminar and turbulent regions. The theoretical
sealing coefficient developed by this a n a l y s i s w a s compared with the
experimental sealing coefficient obtained for ten experimental s e a l s . In
view of the observations and comparisons presented in this work, the
following conclusions are indicated:
1. The laminar a n a l y s i s leading to equation (3), upon which the
present a n a l y s i s i s b a s e d , agrees more closely with the experimental
results in laminar flow than any of the other analytical approaches con-
sidered.
2. The optimum screw geometry for laminar operation was not
optimum for operation in the turbulent r a n g e .
3. For screws which must operate in both the laminar and
turbulent range, smaller helix angles will provide more flexible
operation in both regions .
4. In turbulent operation the visco s e a l s having small helix
angles operate with a greater degree of pressure s t a b i l i t y .
5. Experimental A is e s s e n t i a l l y constant in laminar operation
and in agreement with equation (3).
6. The transition Reynolds number at which A p begins to
d e c r e a s e , d e c r e a s e s with increasing I and is in agreement with
equation (15).
7. The slope din A /dlnRe in the turbulent range for helix
£& C

angles of 9.67® and 14.5° was approximately - 0 . 7 0 , while this value


decreased to - 0 , 4 3 for helix angles of 5.81° .
8. The phenomenon of gas ingestion will be encountered in
turbulent operation. This condition was observed to be a function of
shaft wetability, Reynolds number, seal geometry, and eccentricity ratio,
but until more experimental data are obtained this relationship cannot
be precisely defined. However, in order to minimize the effects of gas
ingestion on A _., large P and ^ along with small a should be employed.
9. There e x i s t s a maximum Reynolds number beyond which long
term operation is no longer f e a s i b l e . However, more experimental data
are needed before this value can be precisely defined.
10. The frictional l o s s e s in the visco seal may be described
with existing e q u a t i o n s . In the laminar range the theoretical predictions
by equation (24) closely approximate the experimental d a t a . During
turbulent operation the experimental data presented by Smith and Fuller,
represented by equation (2 5), are 0.951 times the value of the friction
parameter observed in t h e s e t e s t s . Therefore, the frictional l o s s e s in
the seal may be calculated either for laminar or turbulent operation by:
F.P.g = 4jt ^ /Re ,
and
n 4'^
F.P. = 0.164:t/Re .
o c
11. For laminar operation A p is essentially independent of the
seal eccentricity ratio and under turbulent conditions the effect of seal
eccentricity is slight.
A number of a s p e c t s of the visco seal operation in the
turbulent range require further and more intensive study. Specific
problem areas requiring additional attention are a s follows:
1. More experimental data in the turbulent range for Reynolds
numbers higher than 3000 are needed in order to determine the maximum
Reynolds number for long term seal operation.
2. The construction of t e s t spindles with helix angles of
approximately 5° and P's in the range from 10 to 20 are needed in order
to verify the observations concerning the relatively small effect of P on
J\ _, for high Reynolds numbers and small helix a n g l e s .
3. A refinement of equation (5) for the calculation of A m is
needed in order to better describe the sealing coefficient in the turbulent
region for small helix a n g l e s . It i s realized that the two lines in Figure
3 may not be the best estimate for all geometric parameters.
4. An investigation of the effects of grooves located in the housing
rather than along the shaft would be d e s i r a b l e . This is based on the obser-
vation by Ludwig (4 ) that a smooth shaft located in the threaded housing
would l e s s e n the effects of gas ingestion on A „.
ACKNOWLEDGEMENT
The investigation of dynamic s e a l s was started and initially
supported by The University of T e n n e s s e e . Subsequent support has been
provided by The Oak Ridge National Laboratory and The National Aero-
nautics and Space Administration. The bulk of the work described in
this paper was done under a u s p i c e s of NASA Grant,NsG-587 with Mr. R,
L. Johnson as technical monitor.
5,2.19

NOMENCLATURE
a Axial land width, in.
b ' Axial groove width, i n .
c Radial c l e a r a n c e , in.
D Seal diameter, in.
4jr4
P^p^ _ Re
___E_ Friction parameter
'c
f Resistance coefficient, Figure 3
h Groove depth, i n .
L Active seal lengthy in,
1 Axial threaded length of s e a l , in,
n ^ Number of thread starts
P pressure, Ibf./in.^
q Power l o s s , i n . I b f . / s e c .
Uc/'
l^g _ —j^— Reynolds number based on clearance
t = tan A Tangent of the helix angle
U Surface velocity of the s-eal, i n . / s e c .
X X coordinate in the direction of motion
y y coordinate along shaft axis
z z coordinate in the radial direction
a Helix a n g l e , degrees
h + c
p__ Dlmensionless parameter
^ b
" = ——- Dlmensionless parameter
71 Tl coordinate
A Sealing coefficient a s presented in this work
/^ Absolute v i s c o s i t y , Ibf. s e c . / i n ,
s t coordinate
P Density, Ibf. s e c . ^ / i n . ^
^ Wall shearing s t r e s s , I b f . / i n . 2
Y Dissipation function
Subscripts 1

c Denoting evaluation on clearance


E Denoting experimental
h Denoting film thickness
/ Denoting laminar flow
o Denoting turbulent flow
P Denoting pipe Reynolds number
T Denoting theoretical
n Denoting n direction

! Denoting ^ direction
5.2.19

NOMENCLATURE
a Axial land width, in.
b ' Axial groove width, in.
c Radial c l e a r a n c e , in.
D Seal diameter, in.
4ff A
A3
P^p^ _ Re
_—_c_ Friction parameter
c
f Resistance coefficient, Figure 3
h Groove depth, i n .
L Active seal length, i n .
1 Axial threaded length of s e a l , i n .
n Number of thread starts
P pressure, Ibf./in.^
q P o w e r l e s s , in. I b f . / s e c .
Uc^
l^g _ ___i— Reynolds number based on clearance
c yM
t = tan A Tangent of the helix angle
U Surface velocity of the s e a l , i n , / s e c .
X X coordinate in the direction of motion
y y coordinate along shaft axis
z z coordinate in the radial direction
0 Helix a n g l e , degrees
h + c
p _ — c. — Dimensionless parameter
b Dimensionless parameter
a + b
71 TX coordinate
A Sealing coefficient as presented in this work
/^ Absolute v i s c o s i t y , Ibf. s e c . / i n .
S t coordinate
P Density, Ibf. s e c . 2 / i n . ' *
^ Wall shearing s t r e s s , I b f . / i n . ^
Y Dissipation function
5.2.20

c Denoting evaluation on clearance


E Denoting experimental
h Denoting film thickness
/ Denoting laminar flow
0 Denoting turbulent flow
p Denoting pipe Reynolds number
T Denoting theoretical
H Denoting n direction
1 Denoting J direction
5.2.21

REFERENCES

1. W . K. Stair, "The Visco Seal - A Survey, " Report ME 5 - 6 2 - 2 , The


University of T e n n e s s e e , March 1962, USAEC Report TID - 15987.

2. W . K.Stair and R. H. Hale, "Analysis of the Visco Seal, Part II -


The Concentric Turbulent C a s e , " University of Tennessee,
Mechanical and Aerospace Engineering Research Report ME 66-587-7,
June 2 8 , 1966.

3. W . K. Stair, "Theoretical and Experimental Studies of Visco-Type


Shaft S e a l s , " Semi-Annual Progress Report, University of Tennessee,
Mechanical and Aerospace Engineering Research Report ME 6 6 - 5 8 7 - 5 ,
April 28, 1966.

4. L, P . Ludwig, T.^N. Strom and G. P. Allen, "Gas Ingestion and


Sealing Capacity of Helical Groove Fluid Film Seal (Viscoseal)
Using Sodium and Water As Sealed Fluids, " NASA TN D-3348,
March, 1966.

5. J. M. McGrew and J. D. McHugh, "Analysis and Test of the Screw


Seal in Laminar and Turbulent Operation, " Journal of Basic Engineer-
ing, Trans. ASME, Series D, Vol. 87, 1965, pp. 153-162.

6. W . K. Stair, "Effect of Groove Geometry on Viscoseal Performance,"


to be presented at the ASME Winte-r Annual Meeting, Nov. 27 -
D e c . 1, 1966, New York, N. Y. ASME Preprint 66-WA/FE-28.

7. E. F, Boon and S. E. Tal, "Hydrodynamische Dichtung fur rotierende


W e l l e n , " C h e m i e - I n g . - T e c h n i k . , v o l . 3 1 , no. 3, January 3 1 , 1959,
p p . 2 0 2 - 2 1 2 . Translation by R. Presser published by UKAEA, 1961,
as DEG, Information Series 13 (CA).

8. S. I . Pal, "On Turbulent Flow Between Parallel P l a t e s , " Journal of


Applied M e c h a n i c s , v o l . 20, Trans. ASME, vol. 75, 1953, p p . 109-114.

9. W . K. Stair, "Analysis of the Visco Seal, Part I - The Concentric


Laminar C a s e , " Report ME 65-587-2, The University of Tennessee,
January 18, 1965, NASA Report CR-285.

10. M . I . Smith and D. D. Fuller, "Journal-Bearing Operation at


Superlaminar Speeds, " Trans. ASME, v o l . 78, 1956, p p . 469-474.
A. E, King, "Engineering Information Report on Tests Made With a
Hydrodynamic Seal (Viscoseal) in Oil, Water, and Potassium,"
Westinghouse Electric Corporation, Aerospace Electrical
Division, Report No. WAED 63.3 (Revision A), September, 1964.

Robert L. Johnson, NASA Lewis Research Center, Personal


Communication, August 13, 1965.
TABLE I

DIMENSIONS AND GEOMETRIC PARAMETERS FOR TEST SEALS

D c h a b a
Test Groove Land Groove Angle
Seal Diameter Clearance Depth Width Width Angle (h+c), b
No. Inches Inches Inches Inches Inches Degrees G (a+b)

1 1.2430 0.0042 0.0101 0.0934 0.1595 14.5 3.38 0.631


2 1.2465 0.00235 0.0116 0.1176 0.0494 '9,67 5.94 0.296
3 1.2461 0.00265 0.137 0.0828 0.0842 9.67 6.16 0.504
4 1,2461 0.00265 0.128 0.0514 0.1156 9.67 5.84 0.692
28 1.2420 0.0047 0.00925 0.1176 0.0494 9.67 2.96 0,296
38 1.2408 0.0053 0.01165 0.0828 0.0842 9,67 3.09 0.504
43 1.2408 0.0053 0.01015 0.0514 0.1156 9.67 2.96 0.692
5 1.2455 0.00295 0.0179 0.1585 0.0607 5.81 6.96 0.277
6 1.2455 0.00295 0.0180 0.1070 0.1083 5.81 7.00 0.502
7 1.2455 0,00295 0,0180 0.0609 0.1600 5.81 7,00 0.724
TABLE II
COMPARISON OF TEST RESULTS

T e s t S e a l Number 2B 3B 4B

A ^ (Laminar) 11.73 13.09 12.39 12,86 17.54 14.29 17.20 14.08 12.44 14.10
A £ (Laminar) 10,98 12.86 12.25 11.62 17.90 16.02 14.40 14.92 11.33 11.75

A p
0.93 0.98 0.99 0.90 1.02 1.12 0.84 1.06 0,91 0,83
Ax
(Laminar)
din A E
0.698 0.731 0.695 0.676 0.709 0.709 0.709 0.414 0.430 0.438
dlnRe
(Turb.)
'A
7 L h c ^c
'////A//////A^//////. /

Ui

seal
^"T^ interface a=(|«-y)frD t a n a / n
'D tan «/rH
b= y i r D t a n a/n
^ = ambient pressure

Ps= system pressure

N5
Fig. 1 Basic Elements of a Visco Seal
MTI-2079
A Sealing Coefficient

1-1.
OQ
_ _
.^_^ T ' " T
n
1 1 1 •
o
as
n
en
o 1
O 1
s
j
o 1
i-h
1 Q^
o (
H
n> 1 1
en
ta
3 X
W
s f
00 < m >^
m
H
1 o
c
Q
0
-»•
O <
f\\ ^ m
5 1
0 3- w Q
? 3 J2,. o n
P Q
rt 1 rf" 3
i-h CO
O ^ M »
H 9 \

o - n2 11
O
3
m \ 3"
as
O
en
o
5
O 1 fi Q
W pP J- S
O t
T W 00 Oi 3
o 1 1
5*
1

&. 1

9Z-Z-S
ISJ
Fig. 3 Resistance Coefficient versus Reynolds Number for Pipe Flow
.28

Fig. 4 Theoretical Sealing Coefficient as a Function of p. Re , and 7 for


a = 14.5 , c = 0.003 in., and D = 1.25 in. '^
5.2.29

36

32 / / /

/ /
H 2 8 ^ a = 210.15° U 57 Q P,7°J 5.81°/
z
y
^24
LL
Li.
7
/ / /
LLJ
O20 'T"' "
u 1

/
O 16
y

<
UJ 12
uo
\J><^ ^^
r

< i
Rec = 100
D = 1.250"
1

8 1

1
c = 0.003"
7 = 0.5

0
0 4 6 8 10 12 14 16
P, CLEARANCE RATIO

Fig. 5 Theoretical Sealing Coefficient as a Function of a and p for c = 0.003 in..


D 1.25 in., 7 = 0.5, and Re 100.
A, SEALING COEFFICIENT
^ -^ h^ f^ TO U) U)
O 00 hJ O) O -^ © ?\3 0)

"^"^^
- —
: > - - ^

^yf^

CD >i '• II
O O O r-

O UJ Q
o = ^

11
l\3
P

\ (0
\(J}
\b)
\ ,
-(j|-Va
\
\
V \
CO
CM CD
in
^Y^\°p\ CO
o
o
1 VA-J
CM
N
A
O
II M-l
8 CM oa
\ \ O CO
IW
O DC o
s
o
4J
LLJ o
U c
3

1
PM
o
o ©I to o
o
en tn
cd
II
<
9o o
j j
d o
•S^
1
LJJ
(D o
•r-l TJ
^60 'IJ O M-(
0)
o
CO
«
II II II 4) U Ul
Q uK ^ , / /
^
60O
d
•r4 11
r-l
r
CO K
— ^ 0)
r-l d
CO . H
CM O
•H i n
4 J CM
<u .
U .-4
o
<u II
C D ( D ^ ^ O ( D ( D ^ C\J
O
O
1 N 3 D I 3 3 3 0 3 9NnV3S V 00
•r-l
Ul

10

Fig. 8 Theoretical Sealing Coefficient versus Re for a 5.81° and 20.15° with
7 of 0.3 and 0.7 ^
Ul
Fig, 9 Visco Seal Test Facilities
Oil In /^-Bearing feds

Thrust
Plate

®
-Test Spind le /
® ® ® ®
S^bnt
7 Inlet
^-The rmocouptes Support
Block
Torque Arm •Oil Drain

Fig. 10 Schematic Diagram of the Visco Seal Test Section.


5.2,35

K80 V

^-o- L-75 S
m
1-70 E
•Intercept Pressure

•Supply Pressure

Best line through


points,excluding
»
^^0* supply pressure
a.
2 10-
3
m
m
m

-0A5~*

L
, r 3
Effective Seal Length (in.)

Fig. 11 Typical Pressure and Temperature Gradients in the Visco Seal.


ON

0.6 in./ft. Taper-

Spindles 1-4B
3/16-^ |*-b

•Spindles 5^6^7

Fig. 12 Visco Seal Test Spindles.


1

Ex p e r i m e n t a l C u r v e by Screw N 0.1 c = 0 . 0 0 4 2 ()
M c G r e w & McHugh-y ,— O € = 0.1 0=14.50''
-- — • € = 0.6 0= 3.38
y= 0.631
^ u rP d rs""^ 1 i

10 /
^Laminar
^ - ^
Theory ^ ^ " ^
h%J^<^bx.
' ^T^i^^ i
1
Mctjrew & Mcnugh \f
k'**
\
^

\
45-a. i

\ ^ X

10 10^ 10

Fig. 13 Theoretical and Experimental Sealing Coefficients for Seal 1.


A Sealing Coefficient

^
o

OQ
WJvNO
r°°Ti
D
CO i^
n
*w CO

fD t
O
H 1
fD

J ^ Si
O
JSb
b- 'ch
m XI XI
3 © ©
CL

><!
(D !S
l-i 4i». kdJ

P
rt O O O
P
D
'O b^
O
KO
O
^
f i P
as P ,
o o o ]
CO b b ^
n " o o b
hO NO
l-h CO • • ft
/ •
o o D
J • ^
fD ^/l
p ^w
rf =^n
en
I-h
O
M
CO
fD
1^
^
p
Jp> D
P
a.

1 /
^1

8e'2-g
80 1 _ — « _ ^ — 1
Screw a 0 c
D 5 5.8 V' 6.96 0.277 0.00295"
A 6 5.81^ 7.00 0.502 0.00295"
• 7 5,81 ^ ZOO 0.724 000295"

The o r y
5 nn QCfam
7
6 1^:€^ % ^ ^
\ r u , I
10
^^^i ^
^ . ^ ^ f j bo.
- * T_j

r1^^^
^ \

10^ 10
Re.

Fig. 15 Theoretical and Experimental Sealing Coefficients for Seals 5, 6, and 7.


Ul

MTI-2093
O

Re.

Fig. 16 Theoretical and Experimental Sealing Coefficients for Seals 2 and 2B.
A Sealing Coefficient

O
n>

Co
p

m
xs
fD
H
g-
(D
3

03

p
OP

o
o
m
Ml
[2

ro
P
rt
m
i-ti
O

OS

P
O.
UJ

lf7*Z'e
Fig. 18 Theoretical and Experimental Sealing Coefficients for Seals 4 and 4B.

MTL|209
5
Experiment by
\
Smith & M c G r e w & McH ugh
Fuller -J ^ 1
/ V
ocrew a 0 Y c
2 X D \
o 1 14.50^ 3.38 0.631 0.00420'
.? D 2 9.67*" 5.94 0.296 0.00235"
4 A 3 9.67*" 6.16 0.504 0.00265"
0.10
• \
s^ ^X^^ _. 4 967° 5.84 0.692 0.00265"

1^\ s. _ _.. _.._\.

^ ^ ^
K ng —J
^^^v^S
1^ Smith & Fuller

0.01
\
1 McGrew &
McHugh
^ ^

10' 10
Re,

F i g . 19 T h e o r e t i c a l and Experimental F r i c t i o n P a r a m e t e r s for S e a l s 1, 2 , 3 , and 4.


Re.

Fig. 20 Theoretical and Experimental Friction Parameters for Seals 2B, SB, and 4B.
F i g . 21 T h e o r e t i c a l and Experimental F r i c t i o n P a r a m e t e r s for Seals 5 , 6, and 7.
"^

\ .
SSas—•ss^

\ Of
^
V
1 —

^ ^ ^
1 i 1
\ . — 2E
%^ »
r"^^^
__
1

M n as?s exDerimenia
. . 1

curves IB "X^ X

-A,
~ Scir e w ' a ^ ' y c 5D
h%Sv ^ ^
\J

3C 3.86'' 731 0.50 0.0029"


1B 3.86''IQ15 0.50 0.0020"
or TOXP AfXQ. f% Kr% A A A O c "
—— /
5D 7 26^ 6.31 0.50 aoo 15"

10^ Re. 10^ 10

Fig. 22 Comparison of King's Experimental Data with the Theoretical Sealing


Coefficient
6.1

SECTION 6

TECHNOLOGY III - BEARINGS AND MATERIALS


EXPERIMENTS WITH HYDRODYNAMIC JOURNAL BEARINGS
OF VARIOUS MATERIALS AND DESIGNS IN SODIUM f-i''-
AT TEMPERATURES TO 800° F
by Fredrick T. SchuUer, William J . Anderson, and Zolton Nemeth
Lewis Research Center
National Aeronautics and Space Administration
Cleveland, Ohio
ABSTRACT
Experiments were conducted with 1. 5 inch diameter hydrodynamic
journal bearings in liquid sodium at 500° and 800° F, speeds to 12 000
rpm and unit loads to 31 psi. The stability characteristics of four differ-
ent geometries and the wear and seizure properties of several material
combinations were investigated. Tilting pad bearings were most stable.
Combinations of a cobalt alloy with nickel alloys or with a titanium c a r -
bide cermet showed the best wear and seizure properties.
INTRODUCTION
Extended space exploration missions of the future will necessitate
long periods of continuous, reliable operation of the power generation
system incorporated in the space vehicle. Power levels on the order of
30 000 watts to the million-watt range are anticipated (ref. 1). At the
present time it appears that a turbogenerator system employing a liquid
metal as the working fluid is the most favorable system for the high
power levels desired. Requirements for light weight, high reliability,
and minimum complexity dictate the use of a hermetically sealed pump
with process fluid lubricated bearings to circulate the fluid through the
entire system, including the space radiator.
Fluid film bearings have been selected over rolling element bearings
because the former bearing type maintains a full fluid film more easily.
This eliminates or minimizes the rubbing contact problem which is usually
present in rolling element bearings. Also, the materials problem in fluid
film bearings is much simpler because of less stringent hardness require-
ments.
Fluid film bearings do, however, have some disadvantages over roll-
ing contact bearings, namely, increased power losses, greater breakaway
torques, and the need for a separate bearing to carry thrust loads. In
addition, the disadvantage of principal Interest here is the tendency of
journal bearings to exhibit instability under light or zero load conditions
that will exist in a space vehicle in a zero gravity environment. Instability
here refers to half-frequency whirl or the tendency of the journal center to
orbit about the bearing center at an angular velocity about half that of the
journal around its own center.
Materials chosen as bearing and journal pairs in an alkali metal s y s -
tem must have good resistance to corrosion in the liquid, good wear and
seizure resistance, and a low coefficient ol friction to insure reasonable
breakaway torque, A survey of existing literature on usefulness of mate-
rials for bearings in liquid alkali metals (ref s. 2 to 4) and the need for
using materials of particular expansion coefiicients led to the selection
of the bearing and journal materials reported herein.
The object of this investigation, the data of which a r e reported fully
in reference 5, was to determine the stability characteristics of four
bearing configurations and which bearing and journal material combina-
tions had the best compatability and wear and seizure resistance in liquid
6.

sodium at temperatures to 800 F.


The bearings were submerged in the liquid sodium and operated hydro-
dynamically. Bearings with l-s-inch bore by 1-s-inch length were tested in
liquid sodium at radial loads from 0 to 70 pounds and journal speeds to
12 000 rpm at 500° and 800° F . Bearing friction torque at varying speeds
and loads was recorded and compared with theoretical values under both
laminar and turbulent flow conditions.
APPARATUS
Bearing Test Rig
A cutaway view of the bearing rig, shown in figure 1(a), illustrates
the configuration of the rig and its loading mechanism. The test shaft was
positioned vertically so that normal gravity forces would not act on the
journal. A 15 horsepower d. c. motor powered the test shaft through a
7. 5 to 1 ratio gear box. The sodium test vessel is located immediately
below the main support bearing housing and floats between the upper and
lower gas bearings as shown schematically in figure 1(b). Two semi-
circular wheels connected by a cable belt comprise the radial loading sys-
tem. Radial load was applied by means of an air cylinder between the two
semicircular wheels, one of which pivots on a knife edge. Bearing torque
is measured by a force transducer. Vertical positioning of the test vessel
was achieved by means of a vertically mounted air cylinder located below
the vessel.
The test shaft was mounted on two support ball bearings which were
preloaded to about 40 pounds by means of a wave spring shown in figure 1(a).
This preload was necessary to insure a minimum amount of test shaft run-
out. Cooling fins were mounted on the test shaft immediately below the
.4

bottom support bearing to dissipate heat and prevent excessive lower sup-
port bearing temperatures due to soak back from the high temperature
sodium in the test vessel. The test journal was mounted and keyed to the
bottom end of the test shaft. The test bearing was mounted in a housing
in the test vessel as shown schematically in figure 1(b).
Liquid sodium at 400° F was introduced to the test vessel and heated
to the desired test temperature by means of an induction heater. The
induction heater coil around the test vessel (fig. 1(b)) does not come into
physical contact with the outer surface of the test vessel and therefore
does not inhibit its free swinging motion.
A drain was provided at the bottom of the test vessel to facilitate
draining of contaminated sodium.
Sodium Supply System
The sodium supply system was a noncirculating, once-through system
composed of the following major components: the supply tank, the supply
line, the sodium filter, the filter by-pass line, and the control valve.
The supply tank was sized to contain 20 gallons of sodium and was
equipped with a fill valve, dual thermowells, diffusional cold trap, p r e s -
sure transmitter, vapor trap, and a sodium supply line. The liquid metal
was supplied to the bearing test rig by means of a differential pressure be
tween the tank and the test rig. A diffusional cold trap was provided to
control the sodium oxide content during operation of the sodium system.
The cover gas for pressurizing and venting the supply tank was passed
through a wire-mesh demister (vapor trap) which prevented any sodium
vapors from getting into the cover gas system.
A micrometallic filter was provided in the supply line to the bearing
rig permitting continuous filtering of all supply sodium. The porous
micrometallic filter unit was so designed as to permit replacing the filter-
ing element.
Bearings and Journals
Hydrodynamic bearings of four configurations were tested fully im-
mersed in liquid sodium. Bearings with two axial grooves, three axial
grooves, a herringbone groove journal with a plain bearing, and tilting
pad bearings with three pads were evaluated. One of the three groove
bearings was run with an axial flow pump attached to the test shaft, p r e s -
sure feeding the test bearing through a hole in the test journal.
The bore and length of the bearings in all cases were nominally
% inches. The journal outside diameter and bearing inside diameter
were machined to a 4 to 8 microinch finish, r m s .
A three pad configuration was chosen for the tilting pad bearings
because it affords greater load capacity than a configuration of more than
three pads. Also, it is much easier to accurately aline all pivot points
on the common pivot circle center with a three pad configuration than one
with more than three pads. The load was applied symmetrically between
two support points (pivots).
The bearing and journal materials were: Stellite Star J, Mo-0. 5Ti,
Hastelloy X, Inconel, and titanium carbide (K184B). The composition
and hardnesses of these materials a r e given in table I.
Instrumentation
Dual chromel alumel thermocouples were attached to the test bearing
back and into the liquid sodium in the test vessel. The induction heater
was controlled by one of the dual thermocouples in the sodium bath.
Two capacitance probes, which measured the movement of the test
vessel during a test run, were mounted outside of the test vessel on the
vessel cover, 90° from each other. The signal from the probes was fed
through displacement m e t e r s to an x-y display in an oscilloscope where
the actual pattern of motion of the test vessel could be observed. The
orbital frequency of the test vessel was measured by means of a f r e -
quency counter.
Shorting probes were used in the test vessel to indicate the level of
the sodium. These probes would short out when sodium came into con-
tact with them, thereby either lighting a level indicator light or closing
the main sodium supply valve automatically.
Test shaft speed was measured with a magnetic pickup head mounted
in close proximity to a six-toothed gear on the test shaft. The signal from
the pickup was displayed on a four-channel frequency counter.
PROCEDURE
Pretest Preparation for Two and Three Axial Groove Bearings
P r i o r to each test run, the test bearing was pressed into its housing
with a slight interference fit. The bearing was then machined in place to
a predetermined inside diameter at room temperature that would result in
the desired bore size at test temperature. Nine bore gage readings, each
accurate to within 0. 001 inch, were averaged and used as a measure of the
bearing bore. The outside diameter of the mating journal was then ground
to a size that would result in the desired clearance for the test bearing.
To insure a minimum amount of runout, the journal outside diameter was
machined to within 0.0002 inch concentricity with its inside diameter.
The test bearing housing was then assembled into the test vessel
which was carefully raised into position around the test journal on the
shaft by means of the lower air cylinder.
Pretest Preparation for the Tilting Pad Bearings
The radii of the three pads of a bearing assembly were checked after
delivery from the vendor to insure an accurate geometry. The pads were
assembled into an annular housing by means of a threaded pivot and nut
arrangement. Desired preload was obtained by adjustment of the threaded
pivots until the bearing surfaces of the pads made intimate contact around
a presized set-up plug. The plug was then removed and the tilting pad
housing was assembled in a manner like that for the axial groove bearings.
General Pretest Preparation
After the bearing and journal were assembled and the test vessel
raised to its run position with the upper and lower gas bearings turned
on, the test vessel was filled with alcohol and drained as a final cold
cleaning procedure. The test vessel was then purged with argon and a
cover gas of argon was supplied to the test vessel throughout the test.
The test vessel was preheated to 500° F and liquid sodium at about 400° F
was introduced into the test vessel through the inlet port under about
5 pounds per square inch p r e s s u r e from the 20-gallon supply system.
Sodium flowed into the test vessel until it made contact with the liquid
level probe which automatically closed the main sodium supply valve p r e -
venting overfilling.

With the bearing and journal completely submerged in sodium, the


test vessel was heated to 500° F and allowed to soak for approximately
4 hours to remove remaining contaminants. The vessel was then drained
and refilled with clean sodium and the bearing test was started after equi-
librium conditions had been attained.
Test Procedure
Two types of tests were conducted, one with load and the other at zero
load conditions. Under zero load conditions the semicircular wheel that
pivots on a knife edge was removed along with its loading cable. The force
transducer for torque measurement was attached directly to the semi-
circular wheel on the test vessel.
Shaft speed was increased in 1000 rpm increments from 3000 rpm to
a maximum of 12 000 rpm. In the loaded bearing tests, the loads were
changed a s required to fulfill the purpose of the particular test run. In
some tests the load had to be changed to maintain bearing stability and in
others the load was left unchanged to observe the effect of prolonged insta-
bility in a bearing. The time interval between speed and load changes
varied, but was of sufficient duration so as to allow the friction torque to
stabilize. Speed, load, bearing temperature, and bearing friction torque
were recorded at each speed and load condition.

Test vessel movement was noted by observation of the oscilloscope


screen in an attempt to identify bearing instability at each test interval.
A test was terminated for the following reasons: (a) if the bearing
seized, (b) if the bearing showed indications of imminent failure by e r -
ratic or excessively high torque, (c) if the test vessel no longer floated
freely on its gas bearings, or (d) if the desired objective of the test had
been attained.
RESULTS AND DISCUSSION
Two and Three Axial Groove Cylindrical Bearings
The results of tests on 14 two and three axial groove bearings are
summarized in table II. The results of these tests, which were conducted
over a speed range to 11 000 rpm at unit loads to 26. 7 pounds per square
inch, were generally poor. Some rig problems were encountered in the
early phase of the test program. These problems included test vessel
cocking, which caused excessive wear at the ends of one bearing and two
failures and contamination of the sodium which may have caused four fail-
ures due to particle ingestion into the test bearing clearance area. In
three instances, sodium migrated into the upper gas bearing causing the
test vessel to cock, producing erratic torque readings.
A typical failure caused by contaminant particle ingestion is shown
in figure 2. This type of failure resulted in a rather pronounced gouge in
the bearing and journal. It was after this type of failure that a cleaning
procedure was adopted in which the test bearing and journal were soaked
for 4 hours in the sodium bath at 500° F in the test vessel. The contami-
nated sodium was then drained and a clean supply of filtered sodium was
introduced into the test vessel before a test was begun.

In some instances, at the higher speeds, excessive sloshing of sodium


in the test vessel caused some sodium to migrate up into the top gas bearing
of the test vessel. These sodium droplets would solidify and test vessel
cocking would result causing bearing failure.
The generally poor results obtamed with the two and three axial groove
bearings was principally due to their inherent instability which will be dis-
cussed later.
Herringbone Journal Bearing
Table III summarizes the results of experiments with two herringbone
groove journals running in plain cylindrical bearings. These bearings,
one of which is illustrated in figure 3, were tested in an effort to deter-
mine their stability characteristics.
The two herringbone journal bearings were run over a speed range of
4000 to 12 000 rpm at loads of 4 to 20 psi at sodium temperatures of 500°
and 800° F. Higher bearing torques were observed than for the two and
three axial groove bearings due to the pumping action of the herringbone
grooves.
Bearing MP-3 (table III) ran for 490 minutes and showed very little
wear (fig. 3). The slight scoring of the journal was probably due to a
small contaminant particle. Bearing MP-4 seized after 90 minutes due
to a line heater expansion pushing the test vessel and overloading or cock-
ing the bearing.
Tilting Pad Bearings
Five tilting pad bearings having a three-pad configuration were tested
in sodium at 500° and 800° F at speeds to 12 000 rpm and loads to 70 pounds
(31.1 psi). These results are summarized in table IV. Of the five bearings
tested, one seized due to overloading (bearing number T-2, table IV), and
one seized due to a combination of insufficient clearance and incompatible
bearing and journal material (bearing number T-IA, table IV). The wear
on the unloaded pad of bearings T-1 and T-2A1, recorded in table IV, was
probably due to insufficient preload which resulted in pad flutter.
Table IV lists the various preload coefficients that were investigated.
Although inconclusive, because of the limited number of tests, it appeared
that tilting pad bearings with large machined in radial clearances (0. 0028
to 0. 0036 in.) and high preload coefficients (0. 50 to 0.72) yielded the best
test results, as far a s wear to pads is concerned.
Figure 4 shows a tilting pad bearing that failed due to overloading.
Pads B and C were the loaded pads, with most of the wear shown on the
leading edge of pad C. Analysis of the wear pattern on the outer edges
of all three pads would indicate that the pads had rolled. The surface
damage to the unloaded pad A was attributed to the failure of this bearing
when wear debris was transferred completely around the bearing just prior
to seizure.
Bearing torque data obtained from tilting pad bearings at zero load and
at 10-pound radial load a r e shown in figures 5(a) and (b), respectively.
These data a r e shown compared to the theoretical laminar torque calcu-
lated from the Petroff equation,

T - i^ A ^ L N
^ 120 C^

and the theoretical turbulent torque calculated from the equation given by
Smith and Fuller (ref. 6) where
T^= T^ (0.039 Re^-^"^)
The theoretical torques a r e for a full circular bearing with zero eccen-
tricity, so they a r e only rough approximations for a tilting pad bearing,
and only for zero preload. Bearing torque values indicated that turbulent
flow conditions prevailed over the greater part of the speed range, due
primarily to the low sodium viscosity. The experimental transitional
speed, where the bearing passes from the laminar to the turbulent regime,
occurred at a higher speed than that predicted by theory (critical t r a n s i -
tional speed, Nrp). This may be due to the fact that this transition occurs
over a range of speeds rather than at one definite speed. Similar results
with the transition speed were reported in reference 7.
No abrupt increase in torque readings was noticed when turbulent flow
conditions were approached. The rather gradual increase in torque with
speed made it difficult to specify exactly the speed at which full turbulence
was attained.
Bearing Instability
The bearing instability of principal concern here is half-frequency
whirl. Figure 6 shows oscilloscope traces of bearing motion obtained with
a two axial groove bearing in sodium at 500° F with a 10-pound radial load.
At 4100 rpm the trace indicated stable bearing operation (fig. 6(a)), When
the speed was increased to 5000 rpm, however, the increase in attitude
angle was sufficient to sustain half-frequency whirl. The whirl pattern
observed on the oscilloscope screen is shown in figure 6(b). K the bearing
is allowed to operate unstably, the supporting film between the bearing and
journal soon breaks down, and the bearing eventually fails.
One of the most undesirable characteristics of the two and three groove
bearings was their instability. The result of such instability is graphically
shown in figure 7. Test bearing J - 2 is shown in this figure after 257 min-
utes of operation at 11 psi unit bearing load in 500° F sodium under half-
frequency whirl conditions. The excessive wear shown is the result of
unstable bearing operation.
Of the 14 two and three groove cylindrical bearings tested, five showed
excessive wear due to half-frequency whirl and one seized because of this
instability.
One of the three groove bearings (bearing number M-9, table II) was
loaded sufficiently at each speed throughout its evaluation to keep it run-
ning stably. After 290 minutes of running time in 500° F sodium, the
bearing was removed and no measurable wear was present on either the
journal or the bearing. The maximum load on this bearing was 26, 7 psi
at a maximum speed of 10 000 rpm indicating that a three groove cylin-
drical bearing will run successfully in sodium if properly loaded to sup-
p r e s s half-frequency whirL
Theory indicates that herringbone groove bearings operate at consid-
erably lower attitude angles than do smooth bearings resulting in more
favorable stability characteristics. This type of bearing assembly was
indeed more stable than the two and three groove bearings and plain
journal assemblies. However, the herringbone groove bearing assembly
did show evidence of half-frequency whirl at low load conditions. A more
judicious design of the herringbone groove journal might lead to a bearing
that would be stable even at zero load conditions. Such a bearing has been
run at zero load in air up to 60 000 rpm without any evidence of half-
frequency whirl (ref. 8).
Figure 8 shows the relative stability of the four different cylindrical
bearing configurations tested. The two and three axial groove bearing
configurations were the least stable of the four since they required the
highest load at any specific speed to maintain stable operation. The plain
bearing with a herringbone groove journal was the most stable of the four
since it required the lowest load at a given speed to keep it running stably.
Axial grooved bearings appeared to require a linear increase in load with
speed to maintain stable operation whereas the herringbone groove b e a r -
ing was stable at 25-pound load at speeds of 7000 to 10 000 rpm.
Experimental data on the threshold of instability of two and three axial
groove bearings correlated well with the theoretical curves reported in
reference 9. Figure 9 shows theoretical curves of Sommerfeld number
plotted against the dimensionless critical rotor mass for a 100 partial
bearing and full circular journal bearing. The data points for the two
axial groove and three axial groove bearings generally fall between the
curves indicating good correlation.
The procedure for determining the threshold speed is to calculate the
dimensionless rotor mass

/iDL

Enter figure 9 with this value and determine the corresponding Sommerfeld
number using the appropriate bearing curve. The speed corresponding to
this Sommerfeld number is the threshold speed, that is, the rotor speed at
onset of instabilriy.
The tilting pad bearings were the most stable of the four configurations
tested. However, half-frequency whirl was observed and measured with a
frequency counter on bearing number T-3 at 12 000 rpm and zero load.
Tilting pad bearings T-3 and T-2A1 were run up to 11 000 rpm at zero load
without exhibiting any half-frequency whirl instability.
Although not tested at zero load, the remaining three bearings of the
tilting pad group showed good stability al light loads of 4. 5 psi to speeds
of 11 000 rpm.
Material Compatibility
Table V lists the bearing and journal material combinations that had
good wear and seizure properties in sodium to 800° F.
Stellite Star J material mated with Hastelloy X, titanium carbide
(K184B), or Inconel showed the best wear and seizure properties. Also
titanium carbme |K184B' mated with Mo-0. 5Ti, showed excellent promise.
Materials such as Has.eli'»y X and Inconel having high nickel content,
were prone to catastrophic seizure wh^^n paired in a bearing and journal
combination.
Figure 10 shows the rehults ol a seizure of a three axial groove b e a r -
ing due to an incompatible bectni.g and journal material combination. The
bearing material was Mo-0 5Ti and the journal, Hastelloy X. The surface
of the journal shows a gallmg typical ol this type of failure.
Another example oi the r^sui s of pairing 'ncompatible materials is
shown in figure I L The bearing material was Inconel and the journal
material was Hastelloy X, boih high in mckel content. Seizure resulted
in galling of the pads and iournah The beaiing radial clearance at the
pivots of 0. 0003 inch would not allow many particles to pass through the
bearing without initia+ing suriace damage. With a poor combination of
materials, galling, severe surface damage, and possible seizure quickly
follow the initial surface- damage
Figure 7 shows a*" example ^t manng materials with good compatibil-
ity. The combnid+ic^n ol Stt-'llne Siar J and Hastelloy X material showed
excellent seizure resistance S'nre «his bearing and journal combination
did not seize even aH.er ptol -^ngrd operuion with whirl present which pro-
duced the exce&sivf' w-'ar shov^n in *he figure Figure 12 shows a photo-
graph of the pivot arrangement used on the tilting pad bearings: a sphere
against a flat. Figure 13 shows a typical example of surface damage to
the pivot that was observed in some tests, even after very short runs at
light load. This damage occurred with both the titanium carbide (K162B)
and the Stellite Star J pivot materials which were always mated against
themselves. As shown in table IV, bearing number T-1 ran for 175 min-
utes with a maximum load of 15 pounds on the pivot producing a Hertz
s t r e s s of only 52 600 psi which was sufficient, however, to cause slight
pivot surface damage.
SUMMARY OF RESULTS
A series of hydrodynamic journal bearing experiments was run in
sodium at 500° and 800° F at speeds to 12 000 rpm and loads to 70 pounds.
Four different configurations were tested; cylindrical bearings with two
and three axial grooves, plam cylindrical bearings with herringbone groove
journals, and tilting pad bearings with three pads. The bearing bore in all
cases was 1. 5 inches and all bearings had a length to diameter ratio of 1.
The following results were obtained:
1. The tilting pad bearings were the most stable. Following in order
were: (a) plain cylindrical bearing with a herringbone groove journal, (b)
three axial groove bearing, pressure fed from an axial shaft pump through
a hole in the journal, and 'c'" three and two,axial groove bearings.
2. Stellite Star J material mated with Hastelloy X titanium carbide
(K184B), or Inconel showed the best wear and seizure properties. Also
titanium carbide i'K184B) mated with Mo-0. 5Ti showed excellent promise.
Materials having high n-ickel ronten*-, such as Hastelloy X and Inconel,
6.1,17

were prone to catastrophic seizure when paired in a bearing and Journal


combination.
3. Data on the threshold of instability of circular, grooved bearings
obtained from this investigation compared favorably with predicted theo-
retical values published in reference 9.
4. Actual bearing torque values (for the tilting pad bearings) agreed
favorably with turbulent flow theory, indicating that turbulent flow condi-
tions prevailed over the greater portion of the tested speed range, due
primarily to the low sodium viscosity. Critical transitional speed from
laminar to turbulent flow occurred at a higher speed than the theoretical
Taylor criterion predicts. Similar results were reported in reference 7,
5. No abrupt increase in torque readings was noticed when turbulent
flow conditions were approached, indicating that the transition from
laminar to turbulent flow occurs gradually over a range of Reynolds num-
b e r s rather than suddenly. The rather gradual buildup of torque made it
difficult to specify exactly at what speed turbulence was initiated.
6. Surface damage of the tilting pad bearing pivots was observed in
some tests even after very short runs at light pivot loads. The pivot con-
figuration used in this investigation was a spherical surface against a flat
and the materials were titanium carbide (K162B) against itself and Stellite
Star J against itself.
7. Although inconclusive, because of the limited number of tests, it
appeared as though the tilting pad bearings with large machined-in clear-
ances (0. 0028 to 0. 0036 in. radial) and high preload coefficients (0. 50
to 0. 72) yielded the most favorable test results. (These bearings had
radial clearances at the pivots of 0. 0003 to 0. 0017 in,) The tilting pad
bearings were the most stable of the four configurations tested, however,
bearing instability in the form of half-frequency whirl was observed in
one bearing at 12 000 rpm at zero load conditions,
APPENDIX - SYMBOLS
C radial bearing clearance, in.
D bearing diameter, in,
g acceleration due to gravity, in. / s e c 2
L bearing length, in.
M rotor mass per bearing, W / g , (lb)(sec 2 )/in.
N journal speed, rpm
N' journal speed, rps
Nrp transitional journal speed from laminar to turbulent regime,

41.1 V
TTDC^

R bearing radius, in.


Re Reynolds number, TrDNpC 760//, dimensionless

MN'DL / R V
Sommerfeld number.
w vc^
T^ Petroff's torque for laminar flow, zero load: T, = M7r^D^LN/120C^

T^ turbulent torque: T^ = T^(0. 039 Re°- ^'^) (Smith and Fuller, ref. 6)

W bearing load, lb
W load due to rotor mass, lb (W^ = Mg)
Jc3m
-1— dimensionless critical rotor mass

jU absolute lubricant viscosity, lb sec/in. 2 (reyns)


p kinematic viscosity, centistokes (in. 2/sec)
p lubricant mass density, lb-sec /in.
REFERENCES
1. Slone, Henry O.; and Lieblein, Seymour: Electric Power Generation
Systems for Use in Space. Paper presented at Second Int. Cong.,
Int. Council of Aero Sci., Zurich, Switzerland, Sept. 12-16, 1960.
2. Basham, S. J . ; Stang, J. H.; and Simons, E. M.: Corrosion Screen-
ing of Component Materials for NaK Heat Exchange Systems. Battelle
Memorial Inst., Columbus, Ohio. Preprint 24 Session XXII. Con-
tributed by American Institute of Chemical Engineers, 1958.
3. Apkarian, H.: Investigation of Liquid Metal Lubricated Bearings. Rep.
50GL231, General Electric Co., Nov. 27, 1950.
4. Cook, W. H.: Corrosion Resistance of Various Ceramics and Cermets
to Liquid Metals. Rep. ORNL - 2391, Oak Ridge Nat. Lab., June 15,
1960.
5. SchuUer, Fredrick T.; Anderson, William J . ; and Nemeth, Zolton N.:
The Operation of Hydrodynamic Journal Bearings of Various Materials
and Designs in Sodium at Temperatures to 800° F. (NASA TN to be
published.)
6. Smith, M. I.; and Fuller, D. D.: Journal Bearing Operation at Super-
Laminar Speeds, Trans. ASME, vol. 78, 1956.
7. Stahlhugh, P. H.; and Trippett, R. J . : Liquid Metal Bearing Perfor-
mance in Laminar and Turbulent Regimes. Trans. ASLE, vol. 5,
no. 2, Nov. 1962.
8. Malanoski, S. B . : Experiments on an Ultra-Stable Gas Journal Bear-
ing, MTI-65TR37, Mechanical Technology, Inc., Nov. 1965,
9. Churchill, A, V.: Rotor Bearings Dynamics Design Technology
Part III - Design Handbook for Fluid Film Type Bearings. Tech
Rapt. AFAPL TR-65-45, Part III, May 1965, Wright-Patterson Air
Force Base, Ohio.
6.1.21

o
p-

m
TABLE I. - NOMINAL COMPOSITION AND HARDNESS OF BEARING AND JOURNAL MATERIALS

Materials Rockwell Composition


hardness,
room Al C Or Co Fe Mn Nl Si w Tl TlC Mo Others
temperature

Stellite Star J C-62 .. 2.5 32 40.5 3 .... 2.5 .... 17 ... .. .... 2.5

Hastelloy X B-87 -- 0.2 22 1.5 18 .... 47.2 .... 0.6 . . . .. 9 1.5

Mo-0.5Ti 8-87 „ »-« -- . . . . - .... .... .... .... 0.5 .. 99.5 ...
Inconel B-75 toB-95 -- 0.04 15 .... 7 0.35 78 0.20 .... . . . -- .... ...
Titanium C-67 3 3 .... 40 .... . . . 50 4 "'"
Carbide {K-184B)
(nickel bonded)
E-3570

TABLE n - RESULTS OF 2 AND 3 GROOVE BEARING TESTS IN SODIUM

Bearing Bearmg Journal Journal Bearing Test Measured J o u r n a l speed


Unit load Total test Observed Remarks
number material number material type temper- radial range tested, range tested, time, mstability
ature, clearance rpm psi mm
°F at t e s t
temperature,
m

M-1 M o - 0 5Tl K-6 TiC 2 Groove 500 0.0010 3000 to 8000 4 5 to 2 0 . 0 470 At v a r i o u s speed B e a r m g and j o u r n a l showed m o d e r a t e w e a r
and load conditions due to half frequency w h i r l Increasmg
load stabilized b e a r m g . B e a r m g did not
s e i z e Ran d e l i b e r a t e half-frequency-whirl
operation at m o m e n t a r y i n t e r v a l s

M-7 Mo-0 5Ti K-B TiC 2 Groove 800 0.0010 4000 to 11 000 4 5 to 2 0 . 0 703 None p r e s e n t B e a r m g showed m o d e r a t e w e a r at both e n d s
m d i c a t m g the t e s t v e s s e l had cocked. Bear-
ing did not s e i z e . Sodium had m i g r a t e d to
top g a s b e a r m g of t e s t v e s s e l .

J-1 S t e l h t e Star J HX-B Hastelloy X 3 Groove 500 0 0013 5000 22.0 5 Not s e t up t o Contammant p a r t i c l e s c o r e b e a r m g a n d
observe caused seizure almost immediately after
s t a r t of t e s t

J-2 Stellite s t a r J HX-C Hastelloy X 3 Groove 500 0 0012 5000 to 7000 11 0 257 At 7000 r p m and B e a r m g and j o u r n a l showed s e v e r e w e a r but
11 0 p s i load no s e i z u r e r e s u l t e d . Held b e a r i n g at 7000
r p m and 1 1 0 p s i d e l i b e r a t e l y for 90 m m u t e s
to o b s e r v e r e s u l t s of h a l f - f r e q u e n c y - w h i r l
operation

J-3 Stelhte Star J HX-D Hastelloy X 3 Groove 500 0.0011 5000 to 7000 9.0 260 Not set up to B e a r m g and j o u r n a l showed heavy w e a r
observe probably due to u n s t a b l e operation at light
load B e a r m g did not s e i z e .

HX-B Hastelloy X J-5 Stelhte 3 Groove 500 O.OOU 5000 11.0 30 Not set up to B e a r m g s e i z e d a f t e r 30 m m u t e s a p p a r e n t l y
Star J observe due to c o n t a m m a n t p a r t i c l e .

HX-D Hastelloy X J-6 Stelhte 3 Groove 500 0.0015 3000 to 7000 4. 5 to 8. 9 460 Not s e t up to B e a r m g and j o u r n a l showed heavy w e a r but
Star J observe did not s e i z e Light l o a d s probably c a u s e d
the b e a r m g to r u n unstably.

HX-5 Hastelloy X J-3 Stelhte 3 Groove 800 0.0008 5000 Not r e c o r d e d Not Too s h o r t a run B e a r m g s e i z e d i m m e d i a t e l y , probably due
Star J recorded to o b s e r v e to c o n t a m m a n t p a r t i c l e . B e a r i n g showed a
s c o r e m a r k completely a r o u n d on one end.

^By observation of oscilloscope pattern as the test vessel was moved by hand, to the limits of its outward motion, in a circular orbit about the stationary journal.
E-3570

TABLE 11 - Concluded. RESULTS OF 2 AND 3 GIWOVE BKAMNG TESTS JN SODIUM

Bearmg Bearmg Journal Journal Bearmg Test Measured'' Journal speed Ohitload Total test Observed Remarks
number material number material type temper- radial range tested, range tested, time, mstability
ature, clearance rpm psi mm
°F at t e s t
temperature,
m.

K-3 TiC J-7 Stelhte 3 Groove 500 0.0017 5000 4.5 315 Not set up to B e a r m g and j o u r n a l showed-heavy w e a r but
Star J observe did not s e i z e , B e a r m g apparently w a s
o p e r a t m g unstably b e c a u s e of light load.

M-9 Mo-0.5Ti J-U Stelhte 3 Groove 600 0.0010 5000 to 10 000 1 1 . 0 t o 2 6 . 7 290 At v a r i o u s speed B e a r m g and j o u r n a l showed discoloration
Star J and load conditions but no m e a s u r a b l e w e a r , B e a r m g w a s
purposely loaded sufficient to keep it rvm-
nmg stably. Also r a n d e l i b e r a t e half-
f r e q u e n c y - w h i r l operation at m o m e n t a r y
mtervals.

M-12 M o - 0 5Tl J-1 Stelhte 3 Groove 800 O.OOU 5000 to 10 000 13. 3 to 22. 2 355 None p r e s e n t B e a r m g w a s loaded sufficient to keep it
Star J runnmg stably, however, b e a r i n g s e i z e d at
8000 r p m during shutdown 1/8 mch gall
m a r k a r o u n d b e a r m g possibly c a u s e d by
c o n t a m m a n t p a r t i c l e , o t h e r w i s e light w e a r .

M-10 M o - 0 5Ti J-2 Stelhte 3 Groove 800 0.0010 4000 to 7000 15. 5 t o 1 7 . 8 242 None p r e s e n t B e a r m g s e i z e d at 7000 r p m and 17.8 psi
Star J load. Two s m a l l w e a r a r e a s on each end of
b e a r m g m d i c a t m g the t e s t v e s s e l had cocked.

M-11 Mo-O.STl HX-7 Hastelloy X 3 Groove 800 0.0013 Not r e c o r d e d 0 Not Present B e a r m g s e i z e d i m m e d i a t e l y at 0 load with
recorded immediately defmite mdication of half-frequency w h i r l .
W e a r on both ends of b e a r m g w a s o b s e r v e d .

HX-7 Hastelloy X J-8 Stelhte 3 Groove 500 0.0012 4000 to 9000 1 3 . 3 to 22. 2 647 At v a r i o u s speed A shaft s c r e w pump forced sodium mto the
Star J a n d load conditions b e a r m g through a hole m the j o u r n a l . B e a r -
mg showed s m a l l w e a r a r e a m loaded zone
probably due to d e l i b e r a t e half-frequency-
w h i r l operation at m o m e n t a r y m t e r v a l s .

By observation of oscilloscope pattern as the test vessel was moved by hand, to the limits of its outward motion, in a circular orbit about the stationary journal.
E-3570

TABLE in. - RESULTS OF PLAIN CYLINDRICAL BEARING AND HERRINGBONE GROOVE JOURNAL TESTS IN SODIUM

Bearing Bearing Journal Journal Bearing type Test Measured^ Journal speed Unit load Total Groove angle Number Width Depth Observed Remarks
number material number material temper- radial range tested, range tested, test (measured of of of instability
ature, clearance rpm psi time, from a per- grooves grooves grooves,
°F at test min pendicular and ^ and in.
temper- to the journal lands lands,
ature, axis) ((!) in.
°F

MP-3 Mo-0. 5Ti K-14 TiC Plain bearmg. 500 0.0013 4000 to 12 000 ''O to 17. 8 490 33° 20 0,064 0.0014 At various Light wear on bearing
Herringbone (on lands) speed and and journal, due to
groove journal load con- deliberate half-frequency-
ditions whirl operation at mo-
mentary intervals.

MP-4 Mo-0.5Ti K-A TiC Plain bearing. 800 0. 0008 5000 to 7000 ''O to 20. 0 90 33° 20 0.064 0.0014 At 5000 Bearing seized due to
Herringbone (on lands) rpm and overload caused by
groove journal 0 load heater expansion pushing
test vessel increasing
radial load.
By observation of oscilloscope pattern as the test vessel was moved by hand, to the limits of its outward motion, in a circular orbit about the stationary Journal.
Zero load only momentary since ha If-frequency-whirl occurred.

#
E-3570

TABLE W. - RESULTS OF TILTING PAD BEARING TESTS IN SODIUM

Bearing Bearing Journal Journal Bearing Test Measured^ P r e l o a d J o u r n a l speed Unit load Total Observed Remarks
number material number material type temper- radial coeffi- r a n g e t e s t e d , r a n g e t e s t e d , t e s t instability
ature, clearance cient rpm psi time,
°F at t e s t min
temperature,
°F

T-1 Inconel J-10 Stellite Star J 3 Tilting 500 " 0 ^ = 0.0021 0.19 5000 t o J l 000 4 . 5 to 6 . 7 175 None p r e s e n t Very little w e a r o b s e r v e d with m o s t w e a r on
pad unloaded pad. Pivot s u r f a c e s showed slight
''C = 0.0017 surface damage.

T-2 Inconel J-12 Stellite S t a r J 3 Tilting 500 Cp = 0 . 0 0 1 1 0.36 5000 t o 9000 4 . 5 to 3 1 . 1 311 None p r e s e n t B e a r i n g s e i z e d a t 9000 r p m a t 3 1 p s i b e c a u s e
pad of overloading. Loaded p a d showed m o s t w e a r .
Cp = 0.0007

T-2A1 Inconel J-13 Stellite Star J 3 Tilting 500 C = 0.0028 0.50 3000 to 11 000 0 to 8 . 9 590 None p r e s e n t Unloaded pad showed m o s t w e a r indicating
pad possibly insufficient p r e l o a d . J o u r n a l showed
Cp = 0.0014 light w e a r . P i v o t s showed mating s u r f a c e
damage.

T-3 Hastelloy X J - 1 4 Stellite Star J 3 Tilting 600 C = 0.0036 0.72 2000 t o 12 000 0 to 1 7 . 8 1013 At 12 000 r p m V e r y little w e a r on a l l s h o e s . P i v o t s showed
pad and and 0 load mating surface d a m a g e . C l e a r a n c e a t 800° F
800 Cp = 0.0010
w a s not appreciably different from 500° F v a l u e .

T-IA Inconel HX-8 Hastelloy X 3 Tilting 800 C = 0.0010 0.70 5000 to 8000 4.6 175 None p r e s e n t B e a r i n g s e i z e d when higher load than 4 . 5 p s i
pad w a s a t t e m p t e d , probably due t o tight c l e a r a n c e
C = 0.0003 and incompatible m a t e r i a l s . W e a r evenly d i s -
tributed between p a d s .

By observation of oscilloscope pattern a s the test vessel was moved by hand, to the limits of its outward motion, in a circular orbit about the stationary journal.
C = bearing radial clearance before preload.
*^C = bearing radial clearance at pivot location after preload.
TABLE V. - BEARING AND JOURNAL MATERIAL

COMBINATIONS THAT HAVE GOOD RESISTANCE

WEAR AND SEIZURE IN SODIUM TO 800" F

Inconel

Stellite Star J Hastelloy X

Titanium carbide
(K-184B)

— '••• ——.•..—»••.»••.•.

Titanium carbide Mo-0.5Ti


(K-184B)
6.1.
SPEED PICKUP ^ ^ ^ W A V E SPRING
1 /
i--sr-SUPPORT BEARING
/
SHAFT COOLING FIN"-.^

TEST JOURNAL-^^^ ^ - - ^"^


•^^ *^' ^ - U P P E R THRUST GAS BEARING
TEST BEARING--^
«y'^ ^ - - T E S T VESSEL
LIQUID METAL p T ^ - G A S EXHAUST LINE
FILL TUBE-x^
Uii y / - R A D I A L LOAD CYLIINDER
•JF^% . P < / 4 ^--KNIFE EDGE PIVOT
INDUCTION / FORCE TRANSDUCER
HEATER COIL-
LOWER THRUST
GAS BEARING"-- % |

(a) Cutaway view.


Figure 1. - Liquid metal bearing rig.

LIQUID LEVEL f"


LIQUID 1
METAL
INLET rKNIFE-EDGE
SHAFT ROTAilON / PIVOT
TEST CHA^ b E R — -
:::^^3_
TEST
BEARING
TEST JOURNAL
INDUCTION ^ . ' ct
HEATER ^^•''' /
COIL-'
COUNTERWEIGHT
LIQUID
METAL /
DRAIN-'

CS-41286
(b) Detailed view of test-bearing installation.
Figure 1. - Concluded.
6.1.28 I
•I

C-66-2144
CS-41282

Figure 2. - Seizurecausedby contaminant particle. Bearing (J-1) versus journal (HX-B).

C-66-2149
CS-41284
•P
Figure 3. - Plain bearing (MP-3) versus Herringbone groove journal (K-14) after test.
I 6.1.29

I* JOURNAL

0 INCH 1
1.1II 11111
PAD A PAD B PAD C
C-65-2416
CS-41283

Figure 4. - Failure due to overloading. Tilting pad bearing (T-2) versus journal (J-12).

61—

Tf [0.039 ReO-57]
4 —
ITH AND FULLER)

2 —

UTT^D^LN
(PETROFF)
120 Cr
BEARING •*
TORQUE, .08
( I N . LB)
NSITIONAL SPEED
.06
TTD C ^ V C ,
04

TRANSITIONAL
02 SPEED ( N j )
(TAYLOR
CRITERION)

01
2000 4000 6000 1 0 , 0 0 0 20,000
JOURNAL SPEED, RPM CS-412J
(a) Zero load.
Figure 5. - Comparison of experimental and theoretical friction
torques for a 3-pad tilting pad bearing. Lubricant, 500° F
^ sodium; bearing material, HastelloyX; Journal material,
Stellite Star J; nominal diameter, 1.5 inches; nominal
length, 1.5 inches; radial clearance, 0.0036 inch.
6.1.30
.4

.039 R e ° - " ]
.2
ND FULLER)

.1
.08
BEARING
TORQUE, 06
( I N . LB) • (PETROFF)
.04

ONAL SPEED
V D
o «. V c w «
02

Oil
1000 2000 4000 6000 10,000 20,000
JOURNAL SPEED, RPM
CS-41293
(b) 10 pound load.
Figure 5. - Concluded.

1 CM = 0 . 5 MILS 1 CM = 0.5 MILS


(A) STABLE OPERATION (B) UNSTABLE OPERATION
AT 4100 RPM; 10-LB (HALF FREQUENCY
RADIAL LOAD. WHIRL) AT 5000 RPM;
10-LB RADIAL LOAD.
CS-41290
Figure 6. - Oscilloscope traces of bearing motion obtained with a two-axial groove
bearing in sodium at 500° F. Diametral clearance, 0.0020 inches. Bearing (M-l)
against journal (K-6).
C-65-1556
CS-Aim
Figure 7. - Wear due to unstable bearing operation (one-half frequency whirl). Stellite star
" J " Bearing (J-2) versus Hastelloy " X " journal (HX-C).

60

3-GROOVE BEA

50

2-GROOVE BEARING
r-3-GROOVE BEARING
40
/ PRESSURE LUBRICATED
STABLE
BEARING OPERATION
LOAD, 30 ,-HERRINGBONE
LB GROOVE JOURNAL

20
I/ UNSTABLE
'/ OPERATION

10

2000 4000 6000 8000 10,000


SPEED. RPM CS-41289

Figures - Speed versus load at which bearing half frequency whirl is


initiated Comparison of 4 different bearing and journal configurations.
6.1.32
20,—
I
10 O 3-GROOVE BEARING, 800° F
8 n 3-GROOVE BEARING, 500° F
6 A 2-GROOVE BEARING, 500° F

1
.8 THEORETICAL FULL
CIRCULAR BEARING
.6

THEORETICAL
100 DEG BEARING
2 —

1 ± 111
02 .04.06 .1 .2 4 .6 2 4
HN'DL
S =
W {§;) CS-41287

Figure 9. - Dimensionless critical rotor mass as a function of


Sommerfeld number at the onset of half-frequency whirl for the
plain cylindrical and the 100° partial arc bearings operating with
a turbulent film.

C-66-2148
CS-41281

Figure 10. - Seizure due to incompatible material combination. IV\o -0.5Ti bearing.
•P
(M-11) versus Hastelloy " X " journal (HX-7).

MTI-2070
I 6.1.33

JOURNAL

0 INCH 1
lllllllll
C-66-2151
CS-41279 PAD A PAD C

Figure 11. - Tilting pad bearing seizure due to incompatible material combination. Inconel
bearing (T-IA) versus Hastelloy " X " journal (HX-8).

-FLAT MATING SURFACE

0 1
lllllllll
.—SPHERICAL PIVOT
INCH

CS-41285
Figure 12. - Spfierical pivot and flat mating surface in pad. (Titanium
carbide (K162B)) versus (Titanium carbide (K162B)).

< • 5-41278
SURFACE DAMAGE TO SPHERI
CAL PIVOT. (TITANIUM CARBIDE
(K162B)

Figure 13. - Tilting pad bearing pivot


after operation in sodium.
MTI-2063
ALKALI METAL BEARING & SEAL DEVELOPMENT AT SPACE POWER AND PROPULSION SECTION
by E. S c h n e t z e r
S p a c e Power and P r o p u l s i o n S e c t i o n
Re-Entry Systems Department
M i s s i l e and S p a c e D i v i s i o n
G e n e r a l E l e c t r i c Company
C i n c i n n a t i , Ohio 45215
ABSTRACT
SPPS engaged in three major programs associated with the development of Rankine
cycle power conversion technology for space application: The low viscosity bear-
ing stability investigation under NASA contract, the development of dynamic seals
under Air Force contract and the preparations for a liquid metal bearing test,
simulating the conditions of a space generator, on CIRP funding. All programs
were supported by a corresponding analytical effort, presently geared towards
predicting bearing/rotor response. Experimental results are expected in 1966,
and 1967.

Screw seals, rotating channel and slinger squeeze seals were investigated in the
dynamic seal program. Interface instability was identified as one of the major
problems. It was resolved for the case of the slinger squeeze seal, which was
successfully tested in potassium.

DISCUSSION

Alkali metal bearings and seals comprise two of the basic components necessary
for reliable operation of Rankine cycle turbomachinery for space. As a result
of the high temperatures and radiation fields in nuclear space power systems,
and the general complexity of multi-fluid systems, present turbogenerator
bearings and seals employ the primary thermodynamic fluid of the system in
their operation. SPPS is engaged in the development of alkali metal bearings
and seals for incorporation in a full turbogenerator system.

A typical 350 KWe turbogenerator for space power conversion - General Electric
concept - is shown in Figure 1. It consists of a six-stage turbine driving
an axial gap generator. Each component is supported on two radial bearings
and a thrust bearing. The two components are connected by an elastic coupling.
The design has been optimized for minimum rotor weight and distribution of
this weight over four equal size pivoted pad bearings. Each component,
turbine as well as generator, has its own thrust bearing primarily to fix the
rotor position accurately, in relation to its stator. Thrust loads are mini-
mized in the turbine by proper fluid design, and in the generator by carefully
positioning the rotor in the middle of the stator. The generator cavity is
open to space and is based on the availability of zero leakage seals. The

generator is uncanned and its temperature distribution is so that no potassium

vapor condensation or accumulation of condensate within the generator housing

will take place.

Within SPPS, bearing and seal development is being carried out under

four different programs, three of which are basic, while the last one will

combine the knowledge gained in the three previous ones. Of the three basic

programs, the first one is devoted to the study of bearing stability which

implies the establishment of proper bearing geometries and the verification

of bearing rotor response behavior. This program is carried out under NASA
o
contract and uses water at 120 F as a lubricant (Appendix A and B).

The second program is devoted to the establishment and selection of

suitable bearing materials. Under this program (Contract No. NAS 3-2534,

Appendix C), accurate physical properties have been established for these

materials, such as hot hardness, compressive strength, modulus of elasticity,


o
thermal expansion, potassium wetting and dimensional stability at 1600 F.

Also, corrosion resistance and compatibility in potassium. Friction testing

is being carried out in two highly sophisticated test rigs, one operating in

a vacuum and the other one in a potassium atmosphere. This program favors

extremely hard and stable refractory materials primarily Tungsten and Titanium

Carbides using special binders. This program will not be further discussed

here because of the limitation in time. It covered three years and is close

to its completion, so its results should be available in a reasonable

time.
The third program was carried out under Air Force contract and was

devoted to the development of zero leakage seals employing potassium as a

sealant (Appendix D). Its main application will be in the electric generator

for space.

The fourth program, which is based on the knowledge accumulated in the

three aforementioned programs, is carried out under Company Research Funds

and has as its objective the simulation of the bearing and seal arrangement

in a space generator. This presents the more difficult problem than the

space turbine due to rotor and containment materials involved. This program

will comprise performance and endurance testing of a 70-pound rotor operating

on l^rge carboloy potassium bearings and a zero leakage seal exposed to

vacuum (Appendix E ) .

Bearing Stability Program

The first portion of the Bearing Stability Program was carried out in

1961 through 1963 under NASA Contract No. NAS 3-2111. The test rig in this

time was designed and manufactured by SPPS. Testing and test evaluation,

however, was carried out in the Bearing and Lubricant Center in Schenectady

by Gerry Fox and his staff. The program was continued in 1965 under a new

contract, NAS 3-6479 (Appendix A and B ) .

Under the first contract, six different bearing configurations were

compared in regard to their stability behavior, especially their capability

of avoiding or suppressing partial frequency whirl in the turbulent operating

regime. Out of this program, the pivoted pad and the three-lobe bearing
emerged as the outstanding candidates for stable operation up to high

speeds. These two bearing types are being further investigated under the

present contract under aligned as well as purposely misaligned conditions.

The intent is also to study the effects of seizure of pivoting points on

the four-pad pivoted pad bearing by running tests with one or two pads

fixed.

The test rig for this purpose had to be drastically modified. In

addition, better and more instrumentation has been introduced. Shaft

motions are monitored by four Bently gages per plane, arranged in four

separate planes.

Opposed gages are used combining their signals in summing amplifiers

to avoid errors due to differential thermal expansion. Gages are arranged

on both sides of each bearing to be able to measure shaft bending under

aligned as well as misaligned conditions. As a new venture, provisions

have been made to measure bearing loads and to establish the load vector

by size and angle during testing up to 500 cycles per second. The Bearing

and Lubricant Center has had good success in making such measurements by

using strain gaged buttons for shaft speeds up to 7,000 rpm. For high

speed testing up to 30,000 rpm, originally similar strain gaged buttons

had been considered; however, piezoelectric force gages were finally

selected due to their high spring constant and their extreme capability of

picking up fractions of a pound under preloads of several thousand pounds.

Each bearing support is suspended on Kistler force gages in two planes,

adequate to register the forces transmitted to the bearings.


6

Another innovation is introduced in this test program by using a digital

data handling system for data taking. In high frequency testing, the normal

tools are oscilloscopes and photography, as well as tape recorders. When it

comes to the evaluation of test results, both of these methods are inadequate

and it is usually an extremely time-consuming and difficult process to reduce

tape recordings to printed language. In using a digital data handling system,

special precautions have to be taken in measuring high frequency responses.

This is done by using peak-to-peak detectors for forces and displacement and

average level detectors for equilibrium position of the shaft. The other

inputs, such as speed, torque, flows, temperatures and pressures, are straight

forward. Accurate measurement of the phase angle constitutes a problem in


o
itself and has been accomplished to within + 2 F.

The test rig is shown in Figure 2. It comprises a 1.25-inch diameter

shaft supported by two test bearings approximately 12 inches apart. The shaft

is driven by an electric motor receiving its power from a variable frequency

power supply through a quill shaft to speeds of 30,000 rpm. Side loads are

imposed on the test shaft by two partial-arc loader bearings pneumatically

operated. The lower bearing assembly can be adjusted both transversely and

angularly to allow variation of bearing alignment. Measurements of shaft

motion and force transmitted to the bearings are made in four planes, located

on both sides of the test bearings as shown in detail in Figure 3, Shaft

motion is monitored by Bently inductance gages in an assemblage shown in

Figure 4, The gages have Teflon tips to protect them against water, as

suggested by J. McHugh, To eliminate the effect of shaft inhomogeneities on

gage readings, the shaft is silver plated in the zones of gage measurement
(Figure 5 ) . A measuring accuracy within + 50 micro-inch is an objective

hard to meet. The alignment of the bearings prior to testing is accom-

plished optically by a high-precision telescope in conjunction with optical

targets inserted in the bearings, Figure 6.

Figures 7 and 8 show the test rig under testing conditions at room

temperature (open) as well as temperature controlled for 120 F operation

(closed).

Figure 9 shows the instrumentation and control panel, subdivided into

five compartments, one handling the lube flow, pressure and temperature

control, another for power and frequency remote control of a variable fre-

quency power supply arranged in another building. The third panel holds a

speed counter, the proximity gage oscilloscopes, a temperature recorder and

electronic equipment producing the signals for a digital recording system,

also arranged in a separate building. The fourth panel is devoted to phase

angle and force measurements. The fifth panel holds the equipment required

for vibration and torque readout.

Figure 10 shows the pivoted pad bearing design incorporated in the

present test program. It has four pads with gimbaling points at 55 percent

partial arc. Figure 11 shows the calculated load carrying capacity of this

bearing with oil and with liquid metal as lubricants. This graph brings out

the weakness of the low viscosity bearing resulting in a load carrying capa-

bility of roughly 1/1000 of the oil lubricated bearing.

Computer programs are available which allow prediction of the dynamic

response of a given bearing-rotor configuration. One such program was


6,2

generated by G,E,, Bearing and Lubricant Center, Another one was published

by MTI under Air Force contract, SPPS has used both of these programs and

has generated an additional program which combines VAST and LSE , offering

higher accuracy and flexibility. In the near future, a comparison of actual

test results with those predicted will be possible. Figure 12 shows a typical

predicted curve of bearing stiffness as a function of rotative speed, which

in turn is used to calculate fluid film forces as shown in Figure 13, Figures

14 and 15 show typical amplitudes of shaft motion as a function of speed with

two different levels of shaft unbalance in the easy-fluid test rig, calculated

by using the MTI computer programs. In these curves, the occurrence of critical

speeds can be identified with the high amplitude zones, and it can be seen

that for the larger value of unbalance, the rotor amplitude actually exceeds

the bearing clearance, predicting failure of the bearing. As a major test

objective, generalized dynamic coefficients will be obtained for bearing rotor

system operating conditions involving the practical problems of misalignment

and varying support structure stiffness, both of which are difficult to pre-

duct analytically.

Dynamic Zero Leakage Seal

A program was conducted by SPPS on the development of such a dynamic

seal, Figure 16. Considerable testing in water was done to identify the

proper seal concept, and to prove its hydrodynamic capabilities. Ultimately,

the seal was successfully tested in potassium (Appendix D ) .

VAST - Vibration Analysis Systems


LSE - Linear Simultaneous Differential Equations
Figure 17 shows some of the seal concepts investigated. All of these

share the common characteristic of being non-contacting hydrodynamic seals,

which is a necessary requirement for long-time operation in space turbo-

machinery. Figure 18 shows the physical properties of potassium which

assist the successful functioning of such a seal, where, because of low

vapor pressure, the molecular diffusion from a free liquid surface to the

vacuum of space is small.

A detailed configurational drawing of the dynamic zero leakage seal

ultimately tested is shown in Figure 19. Liquid is injected into the

rotating disc cavity, and sealing is accomplished by centrifugal force of

the liquid. Any migration of liquid along stationary walls toward the

vacuum side results in re-ingestion of the liquid into the seal before it

escapes. Figure 20 is a photograph of the disassembled seal parts. Testing

was accomplished in the liquid metal bearing and seal facility in Figure 21,

which will be used in the forthcoming liquid metal bearing development

program. Figures 22 and 23 show typical performance curves for seals of

diameters of interest.

Figure 24 shows the liquid potassium which was accumulated in a vacuum

tank during 173 hours of testing the dynamic seal. Most of this was accumu-

lated during two startups and shutdowns.

Liquid Metal Bearing Test Program (Appendix E)

The liquid metal bearing test rig, shown in Figure 25, is similar to the

previously discussed water test rig, in that two test bearings support a
6.

vertical shaft driven by an electric motor. The rotor weighs 70 pounds and

simulates the mass and dynamic characteristics of an axial gap generator.

Along with the very accurate manufacturing requirements of the bearing, the

design of a bearing for the generator shaft constitutes a severe mechanical

design requirement, since the rotor of the generator must be made from magnetic

steel, the casing from stainless steel and the bearing is fabricated from

refractory metal carbides having thermal coefficients of expansion of roughly

one-third those of steels. Figure 26. A number of geometric solutions to the

differential thermal expansion problem have been investigated, involving

radial springs, radial faced teeth, and conical faces as shown in Figure 27,

the latter approach having been selected as the most practical. The bearing

detail is shown in Figure 28, and comprises four segmented pads, retained by

spherical pins, allowing self-alignment of the bearing during operation.

The pads are weight balanced so that their centers of gravity coincide with

the pivot contact points to minimize fretting motions. The entire bearing

region is enclosed by close-running shaft screw seals, and is flooded with

potassium. A side load of 200 to 300 pounds can be applied to the rotor

by generating a magnetic field with an induction coil. The test rig has

other new features such as a high temperature induction gage measuring the

potassium film thickness through an aluminum oxide window against an aluminum

oxide ring which is part of the rotor. The lower shaft end is exposed to

vacuum and sealed off by a zero leakage seal modified for a low power require-

ment. Dynamic sealing, in this case, is accomplished by a screw seal. An

additional lift-off face seal provides sealing at standstill and during

startup. Figures 29 and 30 show rotor and stator hardware. The test rig is
presently being assembled and will undergo its easy-fluid checkout in

November 1966. Liquid metal testing is scheduled for the first quarter

of 1967.

Conclusions

The engineering know-how required to resolve the complex problems of a

space turbogenerator operating on liquid metal bearings and employing zero

leakage seals is progressing rapidly. As compared to eight years ago, when

even the problems involved were poorly identified, much knowledge and much

practical experience has been accumulated today. Computer programs are

available and will eventually reach close agreement with actual test condi-

tions, SPPS is prepared to contribute to this general realm of knowledge

by coordinating computer predictions with actual test results, by establishing

best qualified pairs of materials for long-time operation of liquid metal

bearings and by developing seals which will combine the zero leakage feature

with a minimum power consumption.

As a bearing configuration, SPPS presently favors the pivoted pad

bearing not because of its resistance to partial frequency whirl, but also

because of its self-adjustment capability to misalignments or structural

distortions. It is realized that the pivoted pad bearing is complex and

has an inherent weakness of fretting or seizing of pivoting points, A less

complex stabilized sleeve bearing with equal resistance to partial frequency

whirl and supported in a way providing equal self-adjustment to misalign-

ments would be desirable.


6.2.11

The differential thermal expansion problem of the Tungsten carbide

bearing in a steel environment is being actively attacked in our liquid

metal bearing program. Answers and test results should be available in

1967.
APPENDIX A

TITLE OF CONTRACTS JOURNAL BEARING INVESTIGATIONS

NASA CONTRACT NUMBER; NAS 3-2111

DATE CONTRACT WAS ISSUED: OCTOBER 10, 1961

DATE OF FINAL REPORT? DECEMBER 31, 1963

NASA CONTRACTING OFFICE: LEWIS RESEARCH CENTER

NASA PROGRAM MANAGER; JOSEPH P. JOYCE

EXCERPT OF WORK STATEMENT

The Contractor shall conduct an investigation of hydrodynamically


lubricated journal bearings having the capability to suppress shaft whirl
while operating at high journal speeds and with zero bearing load.

The Contractor shall fabricate a test bearing rig with a vertically


mounted shaft (1.25-inch diameter) and speed capability up to 35,000 rpm.

The test rig shall be designed with the facility to alter and to control
accurately the following input variables:

1. Shaft speed
2, Bearing loads
Unit directional
Rotating
3. Inlet lubricant pressure
4. Inlet lubricant temperature
5, Lubricant flow rate

The test rig shall be so designed and instrumented to facilitate the


determination of impending whirl speed and accurate readout of the following
data:

1. Bearing load
2. Inlet lubricant pressure
3. Inlet and outlet lubricant temperature
4. Lubricant flow rate
5. Shaft speed
6. Friction torque
7. Shaft position in the bearing with respect to time
8. Eccentricity
APPENDIX A - (Cont'd)

The Contractor shall study analytically approximately eleven (11)


different bearing configurations and shall then select approximately six
(6) of these for experimental testing.

The Contractor shall determine experimentally the antiwhirl character-


istics for each bearing configuration selected for testing over ranges of
the following variables as specified.

1. Shaft speed: 0 - 35,000 rpm


2. Bearing load: Unidirectional - 0 - 200 lbs.
Rotating - 0 - Maximum level as governed by
maximum attainable speed or by
bearing load capacity
3. Inlet lubricant pressure: 3 - 3 0 psi
4. Inlet lubricant temperature: 100 - 160 F
5. Clearance; 0.0016" - 0,004"
6. Bearing length to diameter ratio (L/D): 0,75 - 1.5

The test lubricant for this test phase shall be distilled water.
APPENDIX B

TITLE OF CONTRACT: INVESTIGATION OF STABILITY OF


HYDRODYNAMIC JOURNAL BEARINGS

NASA CONTRACT NUMBER; NAS 3-6479

DATE CONTRACT WAS ISSUED: APRIL 29, 1965

DATE OF CONTRACT COMPLETION: END OF 1966

NASA CONTRACTING OFFICE: LEWIS RESEARCH CENTER

NASA PROGRAM MANAGER: JOSEPH P. JOYCE

EXCERPT OF WORK STATEMENT

Specific Requirements

The Contractor shall extend the investigation of the stability of the


two hydrodynamically lubricated bearings that have demonstrated under
Contract NAS 3-2111 an ability to suppress half-frequency shaft whirl while
operating at rotational speeds up to 21,000 rpm and with near zero radial
loads. This study shall extend to rotational speeds of 30,000 rpm and shall
include effects of bearing misalignment. The two bearing designs to be
studied are the three lobe and the four pad pivot pad configurations. The
test bearing dimensions are:

Inside Diameter 1.25 inches (nominal)

( Length .
Diameter 1.0

Diametrical Clearance 0.005 inch

The bearing test facility, its auxiliary equipment and instrumentation


used on Contract NAS 3-2111 shall be modified and improved to demonstrate
the capability of testing through the range of variables stated below:

Variables Range

Shaft Rotational Speed 3,600 to 30,000 rpm

2. Bearing Angular Misalignment 0 to 400 sec.


APPENDIX B - (Cont'd)

Bearing Linear Misalignment 0 to 0.004 inches

Combined Bearing Angular and Same as Items 2 & 3


Linear Misalignment

Lubricant Supply Pressure 15 to 150 psia

Bearing Mount Rigidity Rigid 85 50,000 pounds/inch

Bearing Unidirectional Radial Load 0 - 6 0 pounds


(Each Test Bearing)

Bearing Synchronous Rotating Load 0 - 4 0 pounds at 30,000 rpm


(Each Test Bearing)

Bearing Reaction Force Vector 0 to 360


APPENDIX C

TITLE OF CONTRACT: RESEARCH ON MATERIALS SUITABLE FOR


POTASSIUM-LUBRICATED JOURNAL
BEARING AND SHAFT COMBINATIONS

NASA CONTRACT NUMBER: NAS 3-2534

DATE CONTRACT WAS ISSUED: APRIL 22, 1963

DATE OF CONTRACT COMPLETION: END OF 1966

NASA CONTRACTING OFFICE: LEWIS RESEARCH CENTER

NASA PROGRAM MANAGER: ROBERT L. DAVIES

EXCERPT OF WORK STATEMENT

The Contractor shall conduct an evaluation of materials suitable for


potassium-lubricated journal bearings and shaft combinations destined for
use in space system turbogenerators.

The Contractor shall provide capsule corrosion test facilities capable


of subjecting single and/or multiple bearing material specimens to potassium
at 400 F to 1600 F, Test capsules shall be contained in a vacuum better
than 1 x 10~'^tOrr pressure after initial startup and inception of stable
operation of the tests. During the startup time, the vacuum shall be
1 X 10~6torr or better. The capsule facility shall alter, control and give
accurate readout of potassium temperature. Similar facilities shall be
provided to determine dimensional stability characteristics of the bearing
materials.

The Contractor shall provide a physical and mechanical properties test


facilities capable of maintaining a vacuum in the range of lO'^tOrr pressure,
except for thermal expansion measurement apparatus which shall maintain an
inert gas atmosphere. These facilities shall be fitted to determine material
physical mechanical properties.

The Contractor shall provide a friction and wear test facilities capable
of maintaining a vacuum in the range of 10~^torr pressure or better. This
facility shall be designed to alter, control, and read out accurately the
following variables:

1. Test sample sliding speed from 0 to 1000 ft/min.

2. Test sample noimal load to produce yield stress in contact materials.


6.2.17

APPENDIX C - (Cont'd)

3. Induced torque,

4. Test environment
pressure, torr
potassium level.

5. Test temperature 400° to 1200 F.

The Contractor shall provide facilities to evaluate the wetting


behavior of the bearing materials with liquid potassium by the sessile
drop method. The facilities shall be capable of being evacuated to
pressures in the 10~^torr range.

It is anticipated that about 14 materials and about 7 material com-


binations shall be chosen for test.

Corrosion capsule tests shall be conducted as outlined belows

1, At temperatures of 800°, 1200° and 1600°F.

2. Continuous test duration of 1,000 hours at each temperature


listed in 1. above,

3, Capsules shall be fabricated from Cb-lZr Alloy.

4. Test specimen combinations shall be tested with equal


surface areas exposed to potassium.

Dimensional stability - the amount of permanent dimensional change


at constant temperatures shall be conducted at 800°,. 1200°, and 1600°F
for 1000 hours.

The following physical and mechanical properties shall be determined:

1. Coefficient of thermal expansion at 400°, 800°, 1200° and 1600°F.

2. Compressive yield stress at room temperature, 400 , 800 ,


1200° and 1600°F.
o o
3. Ultimate compressive stress at room temperature, 400 , 800 ,
1200° and 1600°F,
o o o
4. Modulus of elasticity at room temperature, 400 , 800 , 1200
and 1600°F.
APPENDIX C - (Cont'd)

5. Hardness at 400°, 800°, 1200° and 1600°F.

Coefficients of friction and wear rates shall be determined under


the following conditions:

1. At temperatures of 400°, 800° and 1200°F.

2. At speeds to 1000 SFM.

3. At loads to 90 percent of the 0.2% yield strength or


ultimate strength of the test materials.
-9
4. At pressures in the 10 torr range and in liquid potassium.
APPENDIX D

TITLE OF CONTRACTS DESIGN CRITERIA-FOR ROTARY SEALS


FOR A SPACE ENVIRONMENT

AIR FORCE CONTRACT NUMBER; AF 33(657)-8469

DATE CONTRACT WAS ISSUEDs MARCH 1, 1962

DATE OF FINAL REPORT: SEPTEMBER, 1965

AIR FORCE CONTRACTING OFFICE: AERONAUTICAL SYSTEMS DIVISION

AIR FORCE PROGRAM MANAGERS ROBERT J. SMITH

EXCERPT OF WORK STATEMENT

Objectives

The objective of this program is to acquire the techniques for com-


pletely sealing high speed rotating shafts under the operating conditions
of high temperature liquid metals and vapors, the near-vacuum environments
of space, and long life.

Requirements

Concepts of seals to meet the objective and the following conditions


shall be designed, fabricated and experimented upon.

1. Fluid - The fluid to be sealed shall be potassium.

2. Temperature - The seals shall be operative at fluid tempera-


tures from the melting point of the fluid selected to 800 P.

3. Pressure - The pressure on the fluid side of the seal shall


be 15 psi and the external pressure shall be 10 mm % .

4. Speed - The speed of the rotating shaft shall be 24,000 rpm.

5. Life - The seal, or seal combinations, shall be designed for


10,000 hours of maintenance-free life.

6. Seal Lubricant - The working fluid, potassium, shall be used


as the seal lubricant.
APPENDIX D - (Cont'd)

7. Leakage - The seal, or seal combinations, shall be capable


of maintaining absolute zero leakage under all conditions
of operation.

8. Size - The seals shall be designed for a 1,0-inch diameter


shaft.

9. Zero G - The seals shall be capable of operation in a zero


"g" environment.

Evaluation

The seal evaluation shall consist of the following:

1. Preliminary experiments

2. 100-hour operational screening t e s t

3. Thermal-cycling t e s t

4. 1000-hour life test

Test Conditions

The test conditions are potassium at 1400 F and 15 psi on one side of
the seal, a pressure of 10" mm Hg on the other side, and a shaft speed of
24,000 rpm.
APPENDIX E

TITIJE; RESEARCH ON LIQUID METAL BEARINGS


AND DYNAMIC SEAL SYSTEMS

FUNDINGS GENERAL ELECTRIC COMPANY


INDEPENDENT RESEARCH FUNDING
(CIRP) DA # L06-09F

PROGRAM INITIATED: 1965

PROGRAM TO BE COMPLETED; MIDDLE OF 1967

Objectives

1. Establish feasibility of potassium lubricated bearings and seals


for application in electric generator.

2. Employ optimum bearing materials as identified by Bearing


Materials Program (NAS 3-2534),

3. Determine bearing stiffness.

4. Establish load carrying capability, dynamic as well as static.

5. Determine power requirements.

6. Determine rotor/bearing response.

7. Establish mechanical integrity

a) During thermal cycling


b) During long-time operation.

8. Establish low power DZL seal.

Specifications

Lubricant Liquid Potassium

Temperature Level i n B e a r i n g s
and S e a l s 600°F

Bearing Diameter 2.625 inches


APPENDIX E - (Cont'd)

Bearing L/D Ratio 1.0

Speed 18,000 rpm

Bearing Type Four-Pad Pivoted


Pad Bearing

Fixed Pivoting Points Preload


Coefficient 0.5

Rotor Weight 70 pounds

Rotor Material H-11 Tool Steel

Residual Unbalance 0.15 gr-in

Provisions for Rotational


Loads up to 1000 pounds By Unbalance

Electromagnetic Side Load Up to 200 pounds

Containment Material 316 Stainless Steel

Bearing Material Carboloy 907


r

n -^

III

Fig. 1. Bearings and Seals in a Potassium Turbogenerator for Space.


ON
m

MTI-2129
6.2.24 I
•I

ElKltlC D i m MOTOI

If,
TOIOUE ftCIUPS ouiit mm
, •• /

UMIAIANCI MiCS

%i TEST SHAfT
EOiCI lUIIOItS
r^

I
lOAOfft l E A i m S
INDUCTANCE 6AUCE DISTANCE OETECTOfiS

I -.jum^-
I

AOJUSIAIIE SUIASSEMIIY
HI »Al

TEST IEAtlN6
» i •» All IN tPROIES)

</
wATti o n ^

F i g . 2. Bearing S t a b i l i t y Test Rig. «P


MTI-2130
6.2.25

Outer Casmq Inner Bearing


Housing

Displacement
Sensor
Holder (4)

iP i (

•I ®

(4)

strumentation / T \
Locations \U

Side Load

Fig. 3. Force Button and Proximity Gage Installation.


t\3

Gage Holder Housing

Probe Lead
i-cickin;.; Cij] 'i!.>t;

jBjMlgfeiJ^

^^H^^<*^
•.

^^^^^?<
-
^^^^^^^^^
^^BP**- -^ Purj^f;
^ ^" , ^' Air In
Air Qui: (Typ . )
Teflon CsTi Pro tec t: i vc Curt a i n

\ \^ s ^ ^.

Fig. 4. Proximity Gage Holder,

(HI-213Z
I ().1.11

Y-J-^fwifl^W*- . .*tT,r?;^'-

Test Shaft
Test Journal Diameter (Lower End)

"titiirrriiiir"-'"f;r '1

;;v:jrfTrjrn|nijii,|r,;[M,j,,pp|if^rf;Mi,n
["•:K.. ° '"'" ^ ^ 3 * 5

Fig. 5 . Test Shaft. MTI-2133

! •

Fig. 6. Optical Alignment Measurement.

MTI-2134
00

ML.
Fig. 7. Test Rig in Room Temperature Test. F i g . 8. Test Rig Temperature C o n t r o l l e d 120 F.

MTI-2135 MTI-2136

% , t. -
SPEED POMER AND
VIBRATIONS PHASE ANRLE SHAFT MOTIONS
TORQUE FREOUENCY
FORCE MEAS. I TEMPERATURES CONTROL

Fig. 9. Instrumentation Panel,


o

1.875"

\-r-\

Fig. 10. Self-Aligning Pivoted Pad Bearing.


•™"'""""'
"~T— _ _ _ j — 1 1 1 1 " k , ^
8000 lOl
«.
m
^^^^'^""^""''^^^ T r
SOOO
?1voted-Pad Bearing
^^x-^-^"^""^ : Shaft Dlaneter, l.ls In.
«noo Kadlal CTearam:e. 1.§ mil
» y^^^^^^^^^-^l Lubricant, Hater IZTF
Preload Coefficient. 0.4

2000 „
, y ^ Shaft Dlaseter » 1.25 In.
"^ C Q . 3 mils

Mm _
800 - _
600 Unidirectional
» ^ Bearing
Stiffness, 1(
400 ^ tfe./ln.
.

200 „ -

100 ™
Radial Displacement » , .^^^^ I
80 Radial Clearance ' ' H ^ ^

60
B B
> / ^ ^ Oj^/^""'""^ I
40

" 10^
y*^ ^ ^ Potassium X. L
I.CbslO* 1.4 1.8 2.2 i.6 3.0
20 -
X totatlve Speed, r^~

10 __l i 1 1 i I 1
4000 MOO 12,000 16,000 20,000 24,M0 28,000
Rotative "Sneed, rpm Fig. 12. Bearing Stiffness Vs. Speed and Side Load.

11. Load Carrying Capacity (Pivoted Pad Bearing)

MTI-2139 MTI-2140 •
200 1
T
Plvoted-Pad Bearing
Shaft Diameter^ 1.25 in.
Lubricant^ Water 120®P
Radial Clearance^ 1.5 mils
Preload Coefficient, 0.4
150 Unbalance5 Upper Bearing 0.50 gr.
Lower Bearing 0,50 gr.
Side Load Per Bearing, 25 lbs.

Maximum
Force, 100
lb.

50

0 10 15 20
R o t a t i v e Speedy rpm x 10"3

Fig. 13. Force Vs. Speed.


6.2.33

— — — Y T" f t r- 1' '" ( "1 1 1 } 1 1 (


\ \ \ \ \ Bearino
\ WSurface
W yys^y^>^v^yys,\v\\-^CV\\\\-X\\\\\\TTVsX\\V\\V^
1.4

"
1.?

1.0 • ""^^^^==rf========,_^^ ;
I.';
irtJfIf^rt^'•

0.6 -
Shdf. DiflBteter. 1.2'j in.
0.4 Ratlial Clearance, 1.5 mil
Unbalance* 0.2S qr. i n .
Side load, 25 lbs.

0.2 .

0 * s * » > S * » « » ' i * * »-^


10 12 14 16 20 22 24 26
Rotativt! Sp^ed. rpm >

F i g - 14» Displacement Vs. Speedy 0.25 i n - g r Unbalance,

MTI-2142

Radial
oisplaces»ntt
nils Squdlbrlijfl
OHplacesent

Shaft Diameter, 1.25 In.


ftadlat Clearance. 1.5 mil
IMbalance, 0.50 q r . - i n .
Sito Load. 25 lbs.

£ 10 14 }6 18 20
Rotative Speed, <

Fig. 15. Displacement Vs. Speed, 0.50 i n - g r Unbalance.

MTI-2143
OBJECTIVE

ESTABLISH "ZERO LEAKAGE" SEAL FOR SPACE GENERATOR

SPECIFICATIONS
SEAL FLUID POTASSIUM AT 600°F
SHAFT DIAMETER 1"
SHAFT SPEED 24,000 RPM
EXTERNAL: LOW VACUUM (lO'S TORR)
LEAKAGE: 1 TO 10 LBS. IN 10.000 HOURS

Fig. 16. Dynamic Seal Development.


Rotating Channel

Vapor
Pressure
(nmHg)
Rotating Dlse

Squeeze Seal

+ &50 700 800 900 1000 1100 1200


Screw Seal Temperature^ Absolute (®H)

Fig. 18. Vapor Pressure of Liquid Metals.


Fig. 17. The Seal Concept.

MTI-2145 MTI-2146
ON

0^

FO".:';N'; »"0.'O'V -jii/c


• : ; > ! : ' • . ! . / '

,-n -•?'•< fit : v i .


; -r-^-?^

i'iacu;.;^^
^.-.iif."

r^OLK^ijifiiS P'jMr-

Fig. 19. Dynamic Zero Leakage Seal.

MTI-2147
r -

t
"N
X
Vsxr^^

Fig. 20. Seal Hardware and Test Spindle.

0^

U3
MTI-2148
0^

00

Fig. 21. Liquid Metal Bearing and Seal F a c i l i t y .

t. .
MTI-2149
6..2.39

Sealing
Differential
Pressure,
psi

10,000 20,000 30.000


Rotational Speed, rpm

Fig. 22. Sealing Capacity of DZL Seal. MII-2150

1 i i — 1 — T " — 1 —

100 ^ J
^1a. » 6 In.

f
80 »- J
/
Conswned
60 ™
/
/
"4
40 —
/
J

20

0
_.

^:+l-
y _ j _
^

_^
Ma. = 4 In.

Dls. »g2 in.


J

10,000 20,000 SO.OM


Rotational Speed, rpm

Fig. 23. Power Consumption of DZL S e a l ,


MTI-2151
ON

to
4>
O

Fig. 24. Seal Leakage in 173 Hours of Testing.


6.2.41

sSv:i'K f-l'/{ i.'-.

(OsfOOi r••?i,^-^

. 'i'.. ,^ »K *! . .• A 'J •;?

•:.':i...^•: .v.: .-;•

s 01)-^

is SH ^ C/U!

i,!J; '.'?' •*'.i A ^ i

i- i. ' i *. 5-. •.

Fig. 25. Liquid Metal Bearing Test Rig.

MII-2153
ON

1 MEAN COEFFICIENTS OF THER. EXPANSION: N3

4>

FOR VARIOUS MATERIALS

m
SI

C
s

TIC^ IO%Cb
s
C A ^ B0LOV B07
3
a- KENNAMlTAL 7170
! •

0 too 200 300 400 500 S0O 7^ BOO


TEMPERATURE *F

Fig. 26. Thermal Expansion Coefficients,


6.2.43

..mmmi marff
CYLINDRICAL ^M^^i^^
© SPRiNe

^ STRAmMT

f^ TAPemo

<^z^zizz:^\
,^ CONICAL I CONTACT ^rm.
^oo mi

'zzzinz:^

Fig. 27. Approach to Differential Thermal Expansion.


6.2.44

•I

K'-^'Hi'.

r-iVOT PIN

i>iVCT PiiiTE

iOUSNAl

BfA?2iHG PAD

'•,

Fig. 28. Carboloy Tilting Pad Bearing.

MTI-2156
r

Fig. 29. Liquid Metal Bearing Test Rig Rotor.

4>
MTI-2157
MOTOR SUPPORT CASING

TOP CASING

MID CASING

BEARING HOUSING
SUPPORT CASING

ROnOM CASING

VACUUM ATTACHMINT
CASING

Fig. 30. Liquid Metal B e a r i n g T e s t Rig S t a t o r .


6 . 3.1-p 1«

Bearing and Seal Materials for Liquid Metal Lubrication

S. Frank Murray

Mechanical Technology Incorporated

ABSTRACT
J ^
The obvious problems of selecting materials for high temperature corrosion
resistance and strength have masked the fact that many of these liquid-metal
lubricated bearing and seal surfaces may fail by seizure long before corrosion
becomes a significant factor. Some of the hard, refractory materials, such as
the cemented carbides, appear to offer a means for achieving good sliding be-
havior, corrosion resistance and high temperature strength all in one package.
Actually, these materials are pacifiers. Design and fabrication problems gen-
erally outweigh the advantages that the cermets offer. In addition, the use
of these materials may limit the size of machine components."— ^

'•^ne promising approach is to use coatings or surface treatments to protect


surfaces in relative motion. The feasibility of this approach has been demon-
strated; however, there is a reluctance to use coatings because of quality
control problems. Unless this problem is resolved, the design and fabrication
of liquid metal lubricated components will continue to be unduly complex and
in some instances, will be a source of built-in unreliability.
INTRODUCTION
There is certainly no controversy over the importance of bearing and seal ma-
terials in process fluid-lubricated systems. In many cases, materials have
been the limiting item in the successful operation of machinery, especially
when liquid metals were used as the lubricating fluids. Their poor viscosity
characteristics, coupled with the fact that liquid metals have little or no
boundary lubricating ability, impose very stringent limitations on the reli-
ability of the bearing systems. Even hydrostatically-supported bearings are
in constant jeopardy of failure because of an accidental rub.

There are materials available with good sliding behavior and corrosion resist-
ance in NaK, but their physical characteristics impose many limitations on
design and fabrication. The purpose of this discussion is to review briefly
the status of the work on bearing materials selection and to recommend an
approach to the problem of achieving good sliding characteristics without
penalizing the bearing or seal design.

BACKGROUND ON BEARING MATERIALS FOR LIQUID METALS


In the early nineteen fifties. Vail, at the Knolls Atomic Power Laboratory and
Basham et al, at Battelle Memorial Institute, both published results of material
compatibility tests with NaK (Refs. 1 - 3 ) . Their results are summarized in
Table 1. The work done at Battelle was concerned with static tests on valve and
seat material combinations in 1500F NaK. The purpose was to determine if weldin
and transfer would take place. Vail's work at KAPL consisted of basic sliding
tests on a large number of material combinations in NaK, at temperatures up to
850F. In both studies, it was found that certain metal-bonded carbides were
very satisfactory. Basham also showed that tungsten and molybdenum were promis-
ing materials for the valve and seat materials.

Since that time, there have been a number of other investigations made to select
bearing material combinations for various liquid metal environments, e.g., Refs.
4-9. In general, the conclusions have been much the same. The cemented car-
bides and tungsten or molybdenum are the most compatible sliding combinations
6

for liquid metal-lubricated bearings. Stellite Star J has also been added to
the list of promising materials. Certainly, these studies were useful since
they broadened our knowledge of the behavior of these materials in various
liquid metals. However, from the standpoint of the design engineer, who is
concerned with building practical machinery, this approach of trying to use
solid carbides or refractory metals appears to be the wrong way to attack this
problem.

That last statement is backed up by practical experience. Some of the early


work on journal bearings operating in NaK was done by Apkarian (Ref. 4) using
the cemented carbides which were recommended by Vail as the bearing materials.
The same background is still being used to select materials, e.g., Figure 1
shows a cemented carbide thrust bearing which was built by MTI for Aerojet-
General under NASA Contract No. NAS 5-417. That bearing ran successfully in
the Snap 8 NaK pump for 3000 hours at temperatures from 300-600F. Pivot dam-
age studies have also been run in 600F NaK using various material combinations,
including the cemented carbides.

Granted that these cemented carbides have excellent sliding compatibility in


liquid metals, there are still many problems involved in the design and fabri-
cation of parts from these materials which discourages their use in machinery.
For example. Table 2 compares some of the characteristics of candidate liquid
metal lubricated bearing materials with common structural alloys. All of the
bearing materials have very low coefficients of thermal expansion and relatively
poor shock resistance. In contrast, most of the structural alloys, with the
exception of the columbium alloy, have much higher coefficients of thermal
expansion. These mismatches in material properties have a strong tendency to
force the design engineer to Ignore sliding compatibility as a bearing material
requirement and to select materials based on other attributes.

For these reasons, MTI has leaned heavily on the use of coatings to provide good
bearing surfaces for unconventional fluid-lubricated systems.
EXPERIENCE ON COATINGS FOR BEARING APPLICATIONS
Most of the work which has been done at MTI on using coatings for bearings has
been directed toward gas-lubricated bearing systems. However, this work does
apply to the problem of liquid metal lubrication. In most of the gas bearing
work, the gases were inert and no contaminating oxide films were present to
provide surface protection. In addition, these non-reactive gases provided no
boundary lubrication. These two conditions apply equally well to lubrication
by NaK or by inert gases, although NaK is certainly a more severe case.

Basically, our approach has been to select one base alloy for the shaft, bear-
ings, and any other critical structural members. This alloy is chosen for
strength, density, corrosion resistance and any other essential attributes.
Then, plasma or flame-sprayed oxide or cermet coatings are used on the journal
and bearing surfaces to provide the necessary resistance to sliding damage.
These coatings are applied, then ground back to a standard thickness of .003
inches. In this way, problems of matching thermal expansion are minimized.
Experience has also shown that these thin coatings are much more capable of
withstanding thermal or mechanical shock than are the solid cermets or ceramics

Even among these so-called wear resistant coatings, there are wide gradations
in effectiveness. Under NASA contract, MTI has evaluated some promising gas
bearing material combinations for use in an inert environment (Ref. 10). In
those tests, a single hydrodynamic tilting pad test bearing was dead weight
loaded down against a 1.5 inch diameter test shaft. A photograph of the par-
tially assembled test rig is shown as Figure 2. Capacitance probes, mounted
on each side of the pad, were used to monitor lift-off speed, running film
thickness and wear. The purpose of those tests was to measure changes in bear-
ing performance as a result of sliding contacts.

In Figure 3, photographs of three different tilting pad bearings are shown


after these bearings had been evaluated in start-stop tests. Two of these
bearings ran very effectively for 1000 starts and stops at a stress level of
4 psi, based on the projected area of the shaft. These were a nickel-bonded
tungsten carbide coated bearing against an Al 0„ coated shaft and a chrome
6

oxide coated bearing against a chrome oxide coated shaft. The contacting sur-
faces were lightly polished and the capacitance probes indicated less than
forty microinches of wear. The third bearing, a hardened M-50 tool steel (60Rc)
running against a cobalt-bonded tungsten carbide coated shaft was effective for
about 350 starts and stops. At this point, the capacitance probe signals in-
dicated that the bearing performance was beginning to degrade. The test was
continued for the full thousand starts and stops, but there the bearing was not
lifting effectively at the end of the test. Examination of the bearing surfaces
after the test showed that the tool steel had transferred and welded to the
carbide coated shaft and that this transferred material was scoring the pad
surface in one area.

These same bearing combinations were also subjected to a series of high speed
rubs by bringing the shaft up to a stable speed of 60,000 rpm and then impacting
the pad against the shaft under various shock loads. Time in contact was on the
order of .02 - .05 seconds. The appearance of the bearings after those high
speed rub tests is shown in Figure 4. The chrome oxide coated pad running
against the chrome oxide coated shaft was still in excellent condition after
ninety rubs. The capacitance probe signals indicated that the bearing was
operating on a larger film thickness at the end of the test than it was in the
beginning. The contacting surfaces were highly polished and smooth.

As was the case in the start-stop tests, the M-50 tool steel pad running against
the cobalt bonded tungsten carbide coated shaft was a very poor combination.
This test had to be stopped before the full sequence of shock loads could be run
because the pad was no longer lifting. The pad was found to be severely cold-
worked and the shaft was covered with transferred metal.

The tungsten carbide coated pad running against an Al 0 coated shaft showed a
gradual deterioration in running performance throughout the test. After ninety
rubs, the probe signals indicated that intermittent contact was taking place
between the pad and the shaft. Examination of the test specimens showed that a
smooth film of the binder metal was smeared on the Al 0„ coated shaft. The
carbide-coated pad was severely abraded, either by the transferred metal or by
carbide grains being torn out of the pad surface.

Similar results have also been obtained with nickel and cobalt bonded tungsten
carbide coated pads running against the same coatings on the shafts.

From the results of these tests the self-mated chrome oxide coating was selected
as being the most promising material for good sliding behavior. This coating has
since been used successfully in a number of gas and water-lubricated machinery
applications.

MTI EXPERIENCE ON COATINGS FOR LIQUID METAL LUBRICATED BEARINGS


When a decision was required on the bearing materials for the NaK lubricated
test rig, which was being built by MTI under AEC Contract No. AT(30-1)3363,
three coatings were selected for preliminary evaluations of their corrosion
resistance to NaK. The selection was based on our previous experience of the
sliding compatibility of these materials in inert environments. These coatings
were:

a) Chrome oxide
b) Chrome oxide, undercoated with molybdenum to improve the adherence of
the oxide.
c) Cobalt-bonded tungsten carbide.

The coatings were sprayed on one end of a hardened, cylindrical, 410 stainless
steel test specimen which was part of a dumbbell tensile test bar. The geometry
and dimensions of the specimens are shown in Figure 5. The coatings were then
ground back to a standard thickness of .003 inches. This specimen configuration
was used so that some measure of the bond strengths of the coatings could be
obtained before and after exposure to NaK. To determine the coating bond strength,
the coated face of the specimen was cemented to the other half of the tensile
test bar with an epoxy resin. The bar was then pulled on a tensile test machine.
Normally, the break will occur at the interface between the coating and the sub-
strate since this bond is rarely as strong as the epoxy joint. Figure 6 is a
6

photograph of some typical test specimens showing the type of coating failure
which should normally occur. By comparing the bond strength of control speci-
mens with the bond strength of specimens which have been exposed to NaK, a
measure of the resistance of the coatings to attack by the liquid metal can
thus be obtained. This bond strength test, coupled with visual observation of
the condition of the coating on the surface, is a reasonable measure of the
resistance of the coating to corrosive attack.

The first test was a 300 hour static test with the specimens immersed in NaK at
200F. The specimens were then removed, examined and subjected to the tensile
test. Table 3 gives the bond strength values for the control specimens (which
had not been exposed to NaK), and for the specimens which had been exposed to
200F NaK. Two discrepancies were noted. First, the bond strength of the chrome
oxide coating, which was undercoated with molybdenum, actually increased by a
significant amount after exposure to the 200F NaK. At this time, we have no
explanation for this behavior.

The second discrepancy was the fact that the joints broke at the epoxy when the
tungsten carbide coated specimens, which had been exposed to NaK, were tested.
At that time, it was felt that this was simply the result of a poor epoxy joint.
However, repeat tests gave essentially the same results and in no case were we
able to get good adhesion between the epoxy and any of the carbide coated speci-
mens after the carbide coatings had been exposed to NaK. It is possible that
the NaK had reacted with the carbide surface or the metal binder to form a film
which did not make a good bond with epoxy. This introduced some uncertainty
into the test results but time did not permit us to resolve this.

There was no visible indication of corrosion on any of the test specimens.

The second part of the screening study consisted of a series of fifty hour tests
in 600F NaK. After each fifty hour cycle, the specimens were removed, washed
and examined. The reason for removing the specimens every fifty hours was to
see if the formation of caustic, as the result of exposure to air, would cause
more corrosion than the continuous immersion in NaK.
After the first fifty hour test at 600F, the chrome oxide coatings were found to
be completely disintegrated. Not only had the coating separated from the surface,
the particles of chrome oxide had also separated from each other and had ended
up in the bottom of the basket as fine debris. The molybdenum undercoat was
still intact on the specimen which had been undercoated. A photograph of one of
the chrome oxide coated specimens after this test is shown in Figure 7.

The cobalt-bonded tungsten carbide specimens had a dark film on the surface, but
the coatings were still intact. These carbide coated specimens were then put
back for another fifty hours. After this second run, there was a faint suspicion
of pitting starting to appear on the surfaces. This could have been due to the
fact that the NaK had washed out some grinding debris which had been embedded in
the surface. In retrospect, this explanation seems reasonable since the pitting
never became any worse.

The tests were suspended temporarily at this point while a new set of specimens
was obtained. This time, 410 stainless specimens with a nickel-bonded carbide
coating (25% WC, 7% Ni + mixed W-Cr carbides) were prepared. This nickel-bonded
carbide is known to be more corrosion resistant than the cobalt-bonded material.
In addition, a tilting pad bearing with this nickel-bonded carbide coating on a
416 stainless substrate was also included. This pad bearing had been ground and
lapped to a 1-3 rms finish for some previous tests. It was included here be-
cause the highly polished surface would make it easier to see if the coating was
actually pitting.

Finally, a type 316 stainless flat with an Al 0 coating was also evaluated just
to see if NaK would attack the Al 0 . The Al 0 coating was as-sprayed.

The specimens described above, plus the two cobalt-bonded tungsten carbide speci-
mens which had already been subjected to the two fifty hour cycles in 600F NaK,
were then run through three more fifty hour cycles at 600F in NaK.

At the end, the specimens were washed and examined. The appearance of the sur-
faces is summarized in Table 4. Figure 8 is a photograph of the polished carbide
6

coated specimen and the Al 0„ coated flat after the test. The carbide coated
pad appeared to be in very good condition, perhaps a little duller than before
the test but certainly not pitted. The Al 0„ coated flat had discolored badly
and looked much rougher. However, the coating was still intact. This test
should be repeated with an Al 0„ coating which has been ground back to a smooth
finish. The result does not prove that Al 0„ is unsuitable, in fact, Al 0„ has
been reported to be resistant to NaK at much higher temperatures but it does
indicate that there maybe some questions as to the ability of the coating to
withstand corrosion by NaKo

The shiny cylindrical specimen, with the threaded end, which is also shown in
Figure 8, is a test sample of 6061 T-6 aluminum alloy that had been plated with
0002 - .0003" of electroless nickel. This aluminum specimen had gone through
the five 50 hour cycle tests at 600F with no apparent corrosion taking place.

Table 5 summarizes the results of the bond strength tests on the carbide-coated
specimens before and after exposure to 600F NaK. Once again, poor epoxy bonds
were obtained on the specimens which had been exposed to NaK, particulary with
the cobalt-bonded carbide. However, the results do indicate that the losses in
bond strength, if any, were not large.

Based on the results of these static corrosion tests, the nickel-bonded tungsten
(W-Cr) carbide coating was selected as being the best choice for the bearing and
journal surfaces in the MTI NaK loop. As far as sliding behavior was concerned,
tbis coating was known to be satisfactory for start-stop operation. Its ability
to withstand high speed rubs was not outstanding, but seemed to be adequate for
several lightly-loaded momentary contacts. The corrosion resistance of the car-
bide coating appeared to be suitable at temperatures up to at least 600F.

This conclusion of the suitability of the carbide coatings for use in 600F NaK
was also shown by Wallace (Ref, 6) who had evaluated cobalt-bonded tungsten car-
bide coatings for their sliding behavior in NaK at 600F, and had found that these
coatings had good wear resistance.
Figure 9 is a photograph of the shaft and pads from the MTI NaK test rig. The
journals and pad surfaces have the nickel-bonded carbide coating. Although the
number of operating hours on this facility is still low, these bearings have
been run in water, alcohol and NaK. As far as sliding behavior is concerned,
the coatings appear to be very satisfactory.

One point which should be emphasized here is the capability of these coatings
for protecting larger areas such as face seal and thrust bearing surfaces. There
is no limit to the diameter of a part as far as the application of coatings is
concerned.

PROBLEM AREAS IN THE USE OF COATINGS FOR PROCESS FLUID LUBRICATION


Up to this point, all of the emphasis has been placed on the attributes of plasma
or flame-sprayed coatings. They do have certain disadvantages, chiefly in the
area of quality control. By varying the techniques used to apply the coatings,
it is possible to get wide variations in coating properties and adherence. This
point was also discussed by Wallace (Ref. 6 ) , who concluded that a vigorous
development effort was required to define a standard coating procedure which
would insure reliable bearing surfaces. Most of the techniques available at
the present time for evaluating the bond strength and integrity of the coatings
are destructive in nature. It is necessary to coat test specimens or coupons
at the same time that the machine elements are being coated and then to use
these specimens for the coating evaluations. This introduces considerable un-
certainty as to the condition of the coating on the actual machine parts. Non-
destructive techniques have been proposed to determine coating adherence. These
include ultrasonics, thermal or infra-red evaluations and eddy current flux
field applications. Each of these is a development problem, but the feasibility
of using each of these techniques has been demonstrated.

There are other problems in selecting the best coating for a given application
and in establishing reasonable temperature limitations for the coatings but this
information can be developed in a straightforward manner.
6.3

CONCLUSIONS
Flame or plasma-sprayed hard coatings offer many significant advantages in sim-
plifying the design and fabrication of process fluid lubricated bearing and seal
components. This Is particularly true for large diameter parts where mismatches
in the thermal expansion coefficients of materials may otherwise pose insur-
mountable difficulties. The effectiveness of certain of these coatings in
protecting the surfaces against sliding damage in inert environments has been
demonstrated in a number of practical applications. All of the test results
which have been obtained to date show that these coatings can also be used
effectively for liquid metal-lubricated systems.

The major drawback to the use of coatings is a lack of adequate quality control.

Ultimately, the choices boil down to these two:

Either we live with the solid cermets and refractory metals and accept the design
problems and limitations which these materials impose, particularly for larger
bearings and seals, or we learn to use coatings and develop the necessary quality
control procedures which can ensure a reliable product.
REFERENCES
1. Vail, D. B. "Compatibility of Materials in Liquid Metal". USAEC Report
KAPL-589. August 18, 1951.
2. Vail, D. B, "Compatibility of Materials in Liquid Metal; Second Report".
USAEC Report KAPL-1021. January 5, 1954.

3- Reactor Handbook. Vol. I - Materials Edited by C. R. Tipton, Jr. Interscience


Publishers, Inc., New York 1960. pp. 762-774,

4. Apkarian, H. "Investigation of Liquid Metal Lubricated Bearings". G.E. Report


R50GL231. November,1950.

5. Kissel, J. W., Glaeser, W. A., and Allen, C. M. "Frictional Behavior of


Sodium-Lubricated Materials in a Controlled High-Temperature Environment".
Wear. Vol. 5 (1962) p 446.

6. Wallace, M. J. "Summary of Potential Liquid Metal Bearing Materials for


SNAP-50/SPUR Pumps". iWAG-468. November 1965. Pratt & Whitney Aircraft-CANEL.

7. Kumpitsch, R. C , Granan, J. R., and Kroon, P. J. "Study of a Liquid Metal,


NaK 77, For Application in Flight Control Systems". ASD-TR-62-597. Vol. 1
May, 1962.

8. Quarterly Technical Progress Report No. APS-5152-R3 SNAP-50/SPUR Contract


by AiResearch Mfg. Co. July 18, 1965. Contract No. AF33(615)2289.

9. Schuller, F. T., Anderson, W. J. and Nemeth, Z. N. "The Operation of


Hydrodynamic Journal Bearings of Various Materials and Designs in Sodium
at Temperatures to 800F". (NASA TN, to be published).

10. Murray, S. F. "Evaluation of Material Combinations for Hydrodynamic Gas-


Lubricated Bearings". Presented at ASME Spring Lubrication Symposium.
New Orleans. June 5-9, 1966.
6.3

CONCLUSIONS
Flame or plasma-sprayed hard coatings offer many significant advantages in sim-
plifying the design and fabrication of process fluid lubricated bearing and seal
components. This is particularly true for large diameter parts where mismatches
in the thermal expansion coefficients of materials may otherwise pose insur-
mountable difficulties. The effectiveness of certain of these coatings in
protecting the surfaces against sliding damage in inert environments has been
demonstrated in a number of practical applications. All of the test results
which have been obtained to date show that these coatings can also be used
effectively for liquid metal-lubricated systems.

The major drawback to the use of coatings is a lack of adequate quality control.

Ultimately, the choices boil down to these two:

Either we live with the solid cermets and refractory metals and accept the design
problems and limitations which these materials impose, particularly for larger
bearings and seals, or we learn to use coatings and develop the necessary quality
control procedures which can ensure a reliable product.
REFERENCES
1. Vail, D, B. "Compatibility of Materials in Liquid Metal". USAEC Report
KAPL-589. August 18, 1951.
2. Vail, D. B. "Compatibility of Materials in Liquid Metal; Second Report".
USAEC Report KAPL-1021. January 5, 1954.

3- Reactor Handbook. Vol. I - Materials Edited by C. R. Tipton, Jr. Interscience


Publishers, Inc., New York I960, pp. 762-774.

4. Apkarian, H. "Investigation of Liquid Metal Lubricated Bearings". G.E. Report


R50GL231. November,1950.

5. Kissel, J. W., Glaeser, W. A., and Allen, C. M. "Frictional Behavior of


Sodium-Lubricated Materials in a Controlled High-Temperature Environment".
Wear. Vol. 5 (1962) p 446.

6. Wallace, M. J. "Summary of Potential Liquid Metal Bearing Materials for


SNAP-50/SPUR Pumps". PWAC-468. November 1965. Pratt & Whitney Aircraft-CANEL.

7. Kumpitsch, R. C , Granan, J. R., and Kroon, P. J. "Study of a Liquid Metal,


NaK 77, For Application in Flight Control Systems". ASD-TR-62-597. Vol. 1
May, 1962.

8. Quarterly Technical Progress Report No. APS-5152-R3 SNAP-50/SPUR Contract


by AiResearch Mfg. Co. July 18, 1965. Contract No. AF33(615)2289.

9. Schuller, F. T., Anderson, W. J. and Nemeth, Z. N. "The Operation of


Hydrodynamic Journal Bearings of Various Materials and Designs in Sodium
at Temperatures to 800F". (NASA TN, to be published).

10. Murray, S. F. "Evaluation of Material Combinations for Hydrodynamic Gas-


Lubricated Bearings". Presented at ASME Spring Lubrication Symposium.
New Orleans. June 5-9, 1966.
TABLE I
mOMISING COMBIMATIONS FOR NaK

S. J. Basham - Battelle D. B. Vail - KAPL


Valve-Seat Tests in 1500F NaK Sliding Tests in 850F NaK

Molybdenum vs. Molybdenum Cobalt-Bonded Tungsten


Carbide vs. Itself
Molybdenum vs. Cobalt-Bonded
Tungsten Carbide Nickel-Bonded Titanium
Carbide vs. Cobalt-Bonded
Tungsten vs. Tungsten Tungsten Carbide

Tungsten vs. Cobalt-Bonded


Tungsten Carbide
TABLE 2
TYPICAL CORROSION RESISTANT BEARING AND STRUCTURAL
MATERIALS FOR LIQUID METAL SERVICE

Coefficient Density - Mechanical


of Thermal Pounds Shock
Expansion , Per Cu.In. Resistance
BEARING MATERIALS in./in./°Fxl0"°
Cobalt-Bonded Tungsten Carbide 2.38 0.54 Poor
Nickel-Molybdenum Bonded Titanium Carbide 3.7 0.22 Poor
Tungsten 2.5 0.70 Fair
TZM (Mo-Base) 2.7 0.37 Fair
Stellite Star J 6.8 0.316 Fa ir-Poor

STRUCTURAL MATERIALS
347 Stainless 9.3 0.29 Good
410 Stainless 5.7 0.28 Good
Columbium - iZr 3.82 0.31 Good
Haynes 25 9.4 0.33 Good
TABLE 3
RESULTS OF BOND STRENGTH TESTS BEFORE AND AFTER EXK)SURE TO 200 F NAK FOR 300 HOURS
ALL TESTS IN DUPLICATE

A. Control Specimens Which Were Not Exposed to Nak

Type of Coating Tensile Stress Average Tensile Percent of Area Broken


at Break Stress Away

Cobalt Bonded Tungsten 3080 psi


3490 psi About 40%
Carbide on 410 Stainless 3900
Chrome Oxide on 1650 psi About 207o
1560 psi
410 Stainless 1470 Less than 10%
Chrome Oxide with 1920 psi Almost complete
1960 psi
Molybdenum Undercoat 2000 separation
on 410 Stainless
B. Specimens After 300 Hour Exposure to Nak at 200F

Cobalt Bonded Tungsten 2175 psi Surfaces did not fail,


2367 psi
Carbide on 410 Stainless 2560 break occurred at__epoxY
Chrome Oxide on 1625 psi About 10%
1662 psi
410 Stainless 1700 Most of coating
Chrome Oxide with 2975 psi Almost complete
3062 psi
Molybdenum Undercoat 3150 separation
on 410 Stainless
TABLE 4
APPEARANCE OF TEST SPECIMENS AFTER STATIC CORROSION TESTS IN NAK AT 600F

Coating After One After Two After Three More 50 Hour Tests
50 Hour Test 50 Hour Tests

Chrome Oxide on Coating


410 Stainless Steel Disintegrated Not Run Not Run

Chrome Oxide plus Chrome Oxide


Molybdenum Undercoat Disintegrated J Not Run Not Run
on 410 Stainless Steel Molybdenum intact

Cobalt-Bonded Tungsten Dark film, Faint suspicion Dark film, easily washed off.
Carbide on 410 easily washed of pitting Same as after first two 50
Stainless Steel off hour tests

Nickel-Bonded Tungsten Covered by dark film, easily


(W-Cr) Carbide on 410 Not Run Not Run washed off. No damage
Stainless Steel visible

Nickel-Bonded Tungsten Covered by dark film, easily


(W-Cr) Carbide on 416 washed off.
Stainless Steel, Lapped Not Run Not Run
Still shows high polish
to High Polish

AI2O0 Coating on 316 Coating still intact but it is


Not Run Not Run
Stainless steel black and appears to be rougher
TABLE 5
RESULTS OF BOND STRENGTH TESTS BEFORE AND AFTER EXPOSURE TO 600F NAK

A. Control Specimens Which Were Not Exposed To Nak

Type of Coating Aver.age Tensile Stre;ss Percent of Area Broken Away

Cobalt-Bonded Tungsten
3490 psi About 40%
Carbide on 410 Stainless

Nickel-Bonded Tungsten (W-Cr) Epoxy bond failed before


>4500 psi
Carbide on 410 Stainless coating

B. Spec imens After Exposure To 600F Nak

Cobalt-Bonded Tungsten Surfaces did not fail,


>2180 psi*^^^
Carbide on 410 Stainless break occurred at epoxy

Nickel-Bonded Tungsten (W-Cr) Surfaces did not fail,


>3800 psi^^^
Carbide on 410 Stainless break occurred at epoxy

(a) After five 50 hour tests


(b) After three 50 hour tests
ON
CO
(-•
00

Fig. 1 Cemented Carbide Thrust Bearing


6.3.19

Fig. 2 Disassembled View of Test Rig Showing


Test Shafts and Tilting-Pad Bearing

Mri-2120
ON

i - 5 0 STEEL CHROME OXIDE

TUNGSTEN CARBIDE
Fig. 3 Tilting Pad Bearings After 1000 Starts and Stops in Argon Atmosphere at 4 psi

H^^Bks
r

M - 5 0 STEEL CHROME OXIDE

TUNGSTEN CARBIDE

Fig. 4 Tilting Pad Bearings After High Speed Rubs at 60,000 rpm in Argon Atmosphere
MTl-1963
6.3.22

HECHANKAl i.^UISOLOGY I'SCORPOHAVETt

TITLE
^ ^ ^ 'r" 1. / -^ /

/ a ^ ^:^£y^£/7^X C^A^y^-^^d^Z

• ,S0€> O/A.
L

7Zy/j scf^/'/9££ ra

J £>^ r^i^^j

UWLESS STANDARD PRACTICES I SURFACES! TOLERANCES


OTHEilWISE t . MALC ! . --ACT O N S I «NGUES

APPHOVCO //.Jo>vej

CHECKED
^//(•/i C j '"»

SCALE
sy/i ^^^3
P R A W N ^ ^ J^^^y^^.«>iiC^^ £^^^

Fig. 5 Geometry of Tensile Test Specimen


ssaniTB^ guiaBoo -jBOTdAx SCTMOHS SBstH-toodg ^sax axrstiax 9 'Sia
..;«=•
6.3.24

Fig. 7 Chrome Oxide Coating After Exposure to 600F NaK

MTI-2122
CZTJ-IIH
in
J009 5B 3BN oa aansodxa ZB^JV snaratoads asax jo aotiBJBaddv g -STJ
CO
•10 Huini'oo o^'TV
S u i a c o o DpTqai?3 (:tD-jvi) u a n s B u n x
popuoq "[3^^TN 'poddr."!
Kv^',^5U^^'%^V'-''^~;,''
UJ

Fig. 9 Shaft and Tilting Pad Bearings from MTI NaK Test
Showing Coatings on Journal and Pad Surfaces

M|[|||24
7.1

SECTION 7

TECHNOLOGY IV - ROTOR-BEARING DYNAMICS


^ BEARING SHAFT SYSTEM DYNAMICS *
MERCURY RANKINE EXPERIENCE AT TRW

by

R. Kasuba
TRW^Inc.
Cleveland, Ohio
SUMMARY

The CRU V's were designed following a combined experimental and analytical
approach. With the past and present general scarcity of rigorous, analytical
methods in deriving a successful mercury lubricated rotor-bearing system
operating beyond the laminar regime, the initially selected analytical tech-
niques were progressively extended or empirically adjusted as needed.

Satisfactory correlations between the predicted and measured responses of the


CRU V rotor-bearing system were obtained.

Deviations between the predicted and measured CRU V rotor-bearing responses


could be further decreased as in most similar complex design cases by:

a) Development of required instrumentation


b) More accurate description of rotor-bearing parameters
c) Improvements of analytical techniques

These further refinements would somewhat improve the accuracy of correlations


between the predicted and experimental rotor-bearing performance; however,
this was not necessarily required for the CRU V development objectives.

The dynamic quantities of the selected three-pad bearing may be characterized


by high damping capacity and wide regions of stable operation.

The CRU V bearing system exhibited the critical damping factors of about .25
to .5 for the considered geometric and operational parameters. As the conse-
quence the shaft orbits and bearing reactions were effectively limited for the
entire speed range and the predicted critical speed could not be experimentally
observed.
The three-pad bearing provided sufficiently large stability regions by select-
ing appropriate pressures and geometric parameters. For example, at the CRU V
operating speed of 36,000 rpm the lubricant supply pressures of 225 psia were
sufficient to prevent fractional whirl instabilities for all variations of
bearing clearances and unidirectional and rotating load combinations.

Finally, the results described here were not only dependent upon development
of improved analytical techniques, but also reflect advanced design and
dimensional control of critical system elements through fabrication and
assembly procedures.
7.1.3

INmODPCTIOM
This presentation s e t s as i t s primary task to discuss b r i e f l y some of the r o t o r -
bearing dynamical q u a n t i t i e s and responses of the Mercury Hankine Power Systems
(SNAP 2) turbo-altornator u n i t s which are being developed for the United States
Atomic Energy Comission by TW. "Hie above turbo-alternator u n i t i , nore popu-
l a r l y named as ttie Combined Rotating Units (CRU's) use a liquid mercury l u b r i -
cated bearing system. A degree of success in developing the CRU's may be
i l l u s t r a t e d by the accumulated operational times. For exanple, the COT ¥ type
r o t o r s have accuinjlated almost 22,000 hours. Individual CRU ? u n i t s show
operational times up to ii700 hours. An additional 20,000 hours were accrued
by Mercury Rankine turbo-Mchineiy u n i t s of e a r l i e r designs.

Design and development of the CRU's was not an ea^y task. The rotor-bearing

system to be considered was in the l i t t l e - k n o w i and investigated nqnlaminar

flow regiiBes involving a f l e x i b l e r o t o r . Stringent bearing require»«nts in

addition to the very limited s t a t e - o f - t h e - a r t involving fluids with .high

i n e r t i a l effects complicated the a n a l y t i c a l design and development of mercury

lubricated bearings. The usual budgetary and engineering schedules for develop-

ing the system were also p r e s e n t . Consequently, within a l l these constraints

the basic CRU rotor bearing system's design was pursued along the p a r a l l e l

and combined paths of a n a l y t i c a l and exj^rimental procedures^

Experimentation and t e s t i n g was conducted a t two levelss

a) Component l e v e l , which was- used for basic development of individual

CEU elements, such as the bearings

b) System level involving the e n t i r e turbo-alternator system with the

preselected bearings.
7.1.4 fl|

In deriving the a n a l y t i c a l design procedures of the rotor-bearing system for

operation in the turbulent r e ^ j i e t h e o r e t i c a l methods were similarly employed in two

stages. For exanple, the r o t o r responses and bearing c h a r a c t e r i s t i c s were

i n i t i a l l y p r e d i c t t d with the available a n a l y t i c a l methods which during bearing

t e s t i n g at the compment level were progressively extended or a n p i r i c a l l y

adjusted as needed^

With the a n a l y t i c a l and experimental techniques used in developing the required

ro,tor-bearing system w e l l defined in relevant references, t h i s presentation

w H l attempt to convey only some of the more tjrpical r e s u l t s or observations

derived from d e s i ^ s of these t u r b o - a l t e r n a t o r u n i t s . The presented r e s u l t s

w i l l primarily include the bearing s t a b i l i t y regions, d i s t r i b u t i o n s of

r o t a t i n g loads, bearing film s t i f f n e s s and damping c o e f f i c i e n t s , and the CRU ¥

rotor-three-pad bearing system responses.


7.1.5

DEVaOPMHT OF THE CRU f ROTOR>BEillIMG SYSTEM

In t h e Mercury Rankine Space Power Systems the s i z e and t h e main mass of t h e

r o t o r and t o some e x t e n t t h e b e a r i n g span are c o n s i d e r a b l y predetermined by

t h e e l e c t r i c a l power r e q u i r e m e n t s j t h u s l e a v i n g the r e s p o n s i b i l i t y f o r o p t i m i z a -

t i o n of t h e dynamical r o t o r - b e a r i n g performance t o t h e proper d e s i ^ of t h e

extending s h a f t s e c t i o n s and the b e a r i n p s . On t h e b a s i s of f o r c e r e s p o n s e

a n a l y s e s , s t r u c t u r a l and t h e r m o e l a s t i c s t u d i e s , some adjustments or trade-offs

were made in d e s i g n s t o co^er t h r e e e n t i r e l y d i f f e r e n t mechanical and t h e r m a l

rep.mes of s t a r t t i p , s t e a d y - s t a t e and spindown. A l a r g e number of c o n s i d e r a -

t i o n s involved in t h e s y s t e m ' s a n a l y s e s may be r e p r e s e n t e d by t h e f a c t o r s

considered in selecting the s u i t a b l e bearings for t h i s specific application.

These f a c t o r s , not n e c e s s a r i l y in the order of importance, may be l i s t e d a s

follows:

Load c a r r y i n g c a p a c i t y

Power l o s s

Bearing f i M c o e f f i c i e n t s t o p r o v i d e s a t i s f a c t o r y range of b e a r i n g -

r o t o r responses

Stability

Resistance to cavitation-erosion

C o m p a t i b i l i t y of r e q u i r e d m a t a r i a l s

Mechanical and t h e r m o e l a s t i c deformations (system and b e a r i n g s )

S a t i s f a c t o r y performance w i t h i n systems p r e s s u r e and flow c a p a b i l i t i e s .

Spinup and spindown t r a n s i e n t c o n d i t i o n s with no m e t a l t o m e t a l c o n t a c t

Operation exceeding Eeynolds numbers of 8,000

Long-term o p e r a t i o n — 10,000 hours

Easy CRU i n t e g r a t i o n
7.1.6

The three-pad, the three-sector, the t i l t i n g - p a d journal bearings, and the

s p i r a l groove and hydrostatic t h r u s t bearings were analyzed for the CHU

application. Employing a systenfs approach which also included t e s t i n g of the

above mentioned bearings, the three-pad journal bearing and the hydrostatic

t h r u s t bearing were selected for the CRU 1 system*

The selected hybrid three-pad bearing shown in Figure ii has t h r e e supply pads
equally spaced around the circumference and three equally sized coitip en sating
orifices. The three-pad bearing can be also described as geometrically
i d e n t i c a l to the three-sector bearing with the exception of the a x i a l drain

grooves.

The main experimental portion of the previously mentioned combined CRU design

program was conducted a t the coitponent t e s t l e v e l . The obtained bearing

data on endurance, s t a b i l i t y and bearing film coefficients provided a d d i t i o n a l

information for selecting the bearings for t e s t i n g at the system's level*

For assuring the dynamical compatibility between these two t e s t l e v e l s the

actual CRU and the counterpart bearing t e s t r o t o r s were designed to be <fynami-


c a l l y similar u n i t s . The mam features of both systems w H l be reviewed by a

few general comments^

CRP V Unit
The CRU ¥ and the associated test setup at the system's test level are illustrated

in Figures 1 and 2. The compactness, electrical, hermetical sealing and

structural requirements limited direct measurements of the shaft orbits at

a single location far outboard of the turbine bearing. Limited direct instru~

mentation and structural inaccessibility made the actual CRU ? somewhat less

suitable for the fundamental and convenient research of the bearings. For
7.1.7

implementation of the shaft orbit data taken at a single axial location, several

accelerometers were installed on the CRU ? housing which was in turn supported

by the test rig structure. Acceleration levels were subsequently used as a

cross-checking element of the calculated shaft orbits for U.e entire rotor

length. The obtained accuracy of correlations of measured orbits and accelera-

tion levels with the predicted counterpart values was quite satisfactory and

will be commented upon in the later stages of this presentation.

The basic instrumentation and display systems used in monitoring the rotor

responses at bo1ii levels of testing included redesigned capacitance orobes

for high temperature application (600^F), panoramic analyzers, dual-beam

oscilloscopes, and magnetic tape recording and playback systems for studies

of startups and shutdowns. The capacitance probe signals were displayed both

as individual probe signals and as Lissajous patterns^ All rotating loads

and dynamic phase angles were referenced to the scribe marks in the shaft at

the capacitance probe locations^

Bearing Test Rig

The basic test rig shown in Fi^nire 3 has capabilities for operation at elevated

temperatures to 500®F« The test rotor which is dynamically similar to the CRU

rotor can be subjected either separately or in combination to both rotating

and unidirectional loads. The unidirectional loads can be introduced in

either upward or downward direction by means of the hydrostatic f^as type load

pads and the desired initial mode of unbalance can be selected by inserting

appropriate unbalancing screws.

The basic instrumentation was similar to that used in the actual CRU's; however,

the structural accessibility to instrumentation permitted the use of four


7.1.8

capacitajxe probes xn sets of two located 90 degrees apart a t two axial loca-
tions. Techniques used in displaying the capacitance probe data a t the system's
level were also employed at t h i s t e s t i n g l e v e l with the sarrple data presented
in Figure 5.
7.1.9

IDADS, BEARDIG STABILITY, AM) BElRBIG F i m COEFFICIENTS OF CRU ROTOR-BEARING


SYSTM

S e v e r a l b a s i c dynamical q u a n t i t i e s a r e needed f o r p r e d i c t i n g the r e s p o n s e s

for a given r o t o r - b e a r i n g system. These r e l e v a n t q u a n t i t i e s g e n e r a l l y i n c l u d e

d e f i n i t i o n of l o a d s , a r e a s of s t a b l e ( n o n - f r a c t i o n a l frequency w h i r l ) o p e r a t i o n

and b e a r i n g f i l m c o e f f i c i e n t s . The p r e s e n t e d q u a n t i t i e s , i t should be men-

t i o n e d , w i l l r e f e r t o t h e CRU ¥ u n i t s and s p e c i f i c t h r e e - p a d b e a r i n g geometries

and t h u s may not be d i r e c t l y a p p l i c a b l e t o o t h e r systems. I t i s hoped, however,

with t h i s p r e s e n t a t i o n to i n d i c a t e the g e n e r a l magnitudes of l o a d s , stability

f i e l d s and t h e range of b e a r i n g film c o e f f i c i e n t s f o r s i m i l a r power systems

i n v o l v i n g mercury l u b r i c a t e d t h r e e - p a d b e a r i n g s o p e r a t i n g in a non-laminar

regime.

Loads and Balancing Techniques

As i n any complex l o n g - d u r a t i o n r o t a t i n g e l e c t r o - m e c h a n i c a l systems t h e CRU

r o t o r s a r e s u b j e c t e d t o b o t h r o t a t i n g and u n i d i r e c t i o n a l l o a d s of v a r i o u s

d i s t r i b u t i o n and sources^ The u n i d i r e c t i o n a l CRU loads c o n s i s t of the r o t o r

w e i g h t , some m a ^ e t i c l o a d s , t u r b i n e n o z z l e and mercury pump l o a d s . These

l o a d s a r e of s i m i l a r magnitudes a s t h e o p e r a t i o n a l r o t a t i n g l o a d s and f o r some

CRU u n i t s t h e y may be s e l f - b a l a n c i n g with t h e r o t a t i n g loads becoming as t h e

primary l o a d i n g mode.

The CRU r o t a t i n g l o a d s a r e comprised of t h e r o t a t i n g magnetic u n b a l a n c e , and

t h e i n h e r e n t and developed mass unbalance l o a d s . The magnitudes of t h e s e

l o a d s were p r e d i c t e d and wherever i t was p r a c t i c a l l y p o s s i b l e they were

measured. For e x a n p l e , t h e m a g n i t u d e s , d i s t r i b u t i o n s and angular o r i e n t a t i o n s

of t h e r o t a t i n g magnetic l o a d s were e x p e r i m e n t a l l y e s t a b l i s h e d w h i l e t h e

d i s t r i b u t i o n of the i n h e r e n t mass e c c e n t r i c i t i e s were d e r i v e d from t h e g e o m e t r i c a l


7.1.10

tolerances. The exact d i s t r i b u t i o n s of the inherent mass unbalances were

not ful3y known; however, t h i s s t i u a t i o n was t h e o r e t i c a l l y analyzed by a s s e s s -

ing the responses for a wide range of p r a c t i c a l and t h e o r e t i c a l d i s t r i b u t i o n s

of inherent mass unbalances.

The approximate range of the CRU ? load magnitudes can be given ass
Rotating magnetic unbalance loads - 2^0 to 7.0 pounds

Direct mass unbalance - mass e c c e n t r i c i t y vectors of about .0002 i n .


All r o t o r s were balanced in two planes within .0003 in-oz at 5600 rpm

Indirect mass unbalance - rotor loads induced due to shaft bending by


various d i s t r i b u t i o n s of mass e c c e n t r i c i t i e s of about 3»0 to 5<.0 pounds
a t 36,000 rpm

Thermally and time induced mass unbalance loads which cannot be p r a c t i c a l l y


counterbalanced at the u n i t ' s assembly were calculated to be approximately
3.0 pounds a t 36,000 rpm

Complex load d i s t r i b u t i o n s in high speed applications always tend to present

a problan in selecting a proper balancing method. Another complication may

arise in t h a t the o f f - s i t e balancing and a c t u a l operational conditions may

be e n t i r e l y different with s t r u c t u r a l i n a c c e s s i b i l i t y permitting no operational

or o n - s i t e balancing.

The multi-plane and two-plane balancing methods were considered and used a t

one or another stage of the r o t o r development program. The type of balancing

and the number of balancing planes can be selected by examining the computed

responses of the blexible r o t o r for various d i s t r i b u t i o n s of the previously

discussed d i r e c t and i n d i r e c t unbalances. Using t h i s t h e o r e t i c a l approach

i t was established t h a t the two-plane balancing a t 5600 rpm w i l l provide for

various load d i s t r i b u t i o n s s a t i s f a c t o r i l y low operational bearing reactions a t

36,000 rpm for the present generations of CRU's.


7.1.11

A characteristic c-inputed response pattern is presented in Figure 12.

Bearing Film Coefficients

The force transmitted into the bearing film or bearing reactions can be

expressed as:

FTRjj = K^ X + C„ X - in the X direction

FTRy = Ky Y + C^ Y - in the Y direction

where K and C are the effective film s t i f f n e s s and darr.ping coefficients in

the given d i r e c t i o n s . Following t l i s representation the q u a n t i t i e s Ky, C^,

Ky and Cy can be experimentally established. As such tl ey are the effective

coefficients which include a l l dynernic and geometric effects of a given

bearing. 'Iheoretically, these effective coefficients can be expressed as

complex functions of e c c e n t r i c i t i e s , d i r e c t and cross-coupled film coefficients

which in t h i s p a r t i c u l a r case must include the significant liquid mercury

inertia effects.

The above effective film coefficients were obtained following hagg and
Sankey from the steady-state solutions of the following basic s e t of equa-
tions of motion;

M X + C^ X + K^ X = F cos (wt + 0)

M Y + Cy Y + ICy Y = F sin (wt + 0)

F and M are the applied unbalance force and the applicable rotor mass per

bearing, respectively, and the mutually perpendicular X and Y axes correspond

to the major and minor axes of the journal locus. Employing tie measured

values of displacements and phase angles of:


I
7.1.12
•I
X = ^ ^ , 0 = 0^ for Y = 0

and

Y = S^i 0 = 0Y for X = 0

the K and C quantities becomes

"cos^X . __...2
+ m U;
h • -^X

v - cos "^1 . mlA;

^sin ^X
c.
X - -J^

^sin ^Y
c.
Y " -JTW

The required displacements and phase angles for solving the above r e l a t i o n s h i p s

were obtained from the component level t e s t s and are i l l u s t r a t e d in Figure $,

Circularity of the Lissajous patterns were maintained u n t i l the u n i d i r e c t i o n a l

loads exceeded the applied r o t a t i o n a l loads by about 1.5 to 2.0 times.

For the case where the predominant loads are r o t a t i o n a l loads such as i s

the case in CRU's the measured o r b i t s are c i r c u l a r and thus the effective

coefficients may be given a s ;

K^ =• Ky and C^ = C^

which then arbitrarily could be referred to as the radial bearing film stiff-

ness and darrping coefficients K and C, respectively. •P

C
7.1.13

Some of the characteristic trends of bearing film coefficients are depicted

graphically in Figures 6 and 7, which show the effects of supply pressures and

two magnitudes of impressed unbalance loads at ambient temperatures for two

clearances which will be designated here as the standard and large (1.3 x

standard) clearances. The basic trends of the hot bearing test results (ijOO^F)

were similar as predicted to the cold test data.

A. For the considered range of loads which represent the eccentricity ratios

below 0.5 the bearing film coefficients of a three-pad bearing can be taken

to be almost independent of applied loads. Most of the spread between the

two load curves can be assigned to a scatter of experimental data points.

B. The bearing film stiffness curves suggest again two distinctive regions

of operation of these hybrid bejrings. In the low speed region the

hydrostatic action appears to be the predominant mode wl ich is identified

by somewhat flattened stiffness curves. In this region the bearing film

stiffness is strongly dependent on the magnitude of supply pressure. At

the inception of the high speed region somewhere above 25,OOU rpm range

the hydrodynamic action begins to predominate and the stiffness becomes

more dependent on speed than on the initial supply pressure.

The general trends of the obtained bearing film coefficients of a three-

pad bearing may be identified as:

Film stiffness - increasing slope with speed

Film damping - somewhat decreasing slope with speed to flattened

curves for some geometric bearing properties

C. The thermal tests suggested the following adjustments:

1) The bearing film stiffness is generally reduced from 0 to lS%

depending on the speed by increasing the lubricant supply tempera-

ture from 70° to ijOO^F. VJithin the same temperature range the
7.1.14

bearing film damping is reduced by approximately 26% which is

equivalent to the percentage reduction in mercury viscosity.

Stability Threshold

Two stability threshold patterns of bearing supply pressure versus shaft speed

with no load, and unidirectional load versus bearing flow at a given speed are

given in Figures 8 and 9« These threshold stability patterns were derived

experimentally for the vertical and horizontal shaft operation and can be

considered as the limiting patterns for the system since they represent the

largest expected bearing clearances.

The instabilities observed vrith the three-pad bearings for several orifice

sizes did not exhibit an abrupt transition between stable and unstable opera-

tion. The whirl orbit of these bearings gradually increased at the onset of

instability for each incremental parameter change until reaching a stabilized

orbit plateau when the bearing was operated viell into the unstable region.

This transitional area may be arbitrarily defined as the region of hysteresis.

The orbits of these bearings in the instability region remained several

times below the available bearing clearance. In characteristic unstable

gas or oil lubricated bearings the whirl orbits typically increase from a

stable small orbit quite rapidly to a whirl orbit approximating the bearing

clearance orbit. The observed fractional frequency whirl was at .[48 to .Ii9

of the shaft whirling speed.

The fractional whirl instabilities were easily identifiable by harmonic

analyzers and also by four notches in the Lissajous pattern or two notches

in individual traces of the capacitance probe data. Otherwise the recorded

data vras similar to one shown in Figure 5.

- \
7.1.15

In summary, a t the CRU V operating speed of 36,000 rpm the lubricant supply
pressures of 225 p&ia were sufficient to prevent fractional whirl i n s t a b i l -
i t i e s for a l l operational v a r i a t i o n s of bearing clearances and unidirectional
and r o t a t i n g load combinations.
7.1.16

RESPONSES OF THE CI-U V ROTOR-BEARING SYSTEM

The main o b j e c t i v e s d u r i n g t h e system l e v e l o r CRU V t e s t i n g were d i r e c t e d

toward c o r r e l a t i o n of the meas-ured and p r e d i c t e d s h a f t o r b i t s . This was an

imposing t a s k s i n c e only one a x i a l l o c a t i o n was a c c e s s i b l e in t h e CRU f o r

i n s t a l l a t i o n of t h e c a p a c i t a n c e p r o b e s .

R e f e r r i n g to the p r e v i o u s l y reviewed dynamic q u a n t i t i e s , some of t h e t y p i c a l

CRU ? r o t o r - b e a r i n g system r e s p o n s e s w i l l be d i s c u s s e d . The i n i t i a l b a s i s

of d i s c u s s i o n w i l l be s e t by e s t a b l i s h i n g t h e t h e o r e t i c a l r e l a t i o n s h i p between

the o p e r a t i o n a l and c r i t i c a l speeds of t h e system. As shown i n F i g u r e 1 0 ,

the i n t e r s e c t i o n s of the s t i f f n e s s c u r v e s w i t h the c r i t i c a l speed curves

i n d i c a t e t h e CRU ? o p e r a t i o n between t h e f i r s t and second c r i t i c a l speeds

with t h e p a s s a g e through t h e f i r s t c r i t i c a l speed between 17,000 t o 27,000 rpm

depending on b e a r i n g c l e a r a n c e s and o o e r a t i o n a l c o n d i t i o n s .

The CRU ¥ r o t o r - b e a r i n g system i s a h e a v i l y danped ( c r i t i c a l damping f a c t o r

of .25 t o . 5 ) and c o m p a r a t i v e l y l i g h t l y loaded system and, consequently,

t h e dynamical r o t o r o r b i t s and b e a r i n g r e a c t i o n s were e f f e c t i v e l y limited for

the e n t i r e o p e r a t i o n a l speed r a n g e . Experimental c o n f i r m a t i o n of critical

speeds b a s e d on a c c e l e r a t i o n l e v e l s and c a p a c i t a n c e probe measurements, i n c l u d

ing spindowns and s t a r t u p s , could n o t be f i r m l y e s t a b l i s h e d , although t h r e e

types of i n d i r e c t checks were a v a i l a b l e to e s t a b l i s h the CRU ¥ c r i t i c a l speed

ranges:

A. Transient s t a t e - startups and spindowns

B. Steady-state operation with a fixed set of parameters a t various

speed plateaus

C. Steady-state speed ooeration with varying parametric inputs to affect

bearing c h a r a c t e r i s t i c s .
7.1.17

With orbit measure! ^nts taken at a single axial location far outboard of the

turbine bearing, thr; maximum orbit at that location for some load configura-

tions may not determine the actual critical soeed. The phase angles between

the rotating forces and displacements at critical speeds easily visualized

for the equivalent rigid maps type members - aisplacement relationships are

not directly transferable to the flexible tjne lumbers under various load

distribution. Finally, with a small change in transmissibility ratios of

about 20^ or luwer for the entire CRU speed range, the small changes in

acceleration at low acceleration levels are hardly detectable.

Closest experimental confirmation of the CRU V critical speeds can be based on

the component level test data which indicatestinefirst critical speed for

the dynamicu.iiy equivalent CRU V shaft in the predicted critical speed ranges

defined by a change in phase angle of 90 degrees. Test data at the component

level test were characteristic of a heavily damped system with no dramatic

changes of orbits and phase angles between the aoplied shaft forces and

journal displacement vectors over the entire speed range from 0 to 145,000 rpm

or above.

The behavior of rotor whirl orbits in the entire operational CRU V speed

range is shown in Figure 11, This presentation contains two types of datat

a) Envelope of measured and calculated startup and suindown shaft

extension orbits for a large number of CliU ? units.

b) Calculated journal orbits representing the more heavily loaded CRU V

units at average operating conditions.

These two quantities are presented in relationship to the standard bearing

clearance to indicate an approximate degree of available load capacity.


7,1.18

In subsequent part, aphs the developed bearing reactions w i l l be correlated

with the predicted .-^nd measured acceleration levels*

The approximate d i s t r i b u t i o n s of tJie previously discussed loads are shown in

Figure 12 with the t h e o r e t i c a l shaft o r b i t s calculated at the design speed

of 36,000 rpm for a t^/pical set of 'oesrtnp, ot'^'na''iwnai c.-)nditions (approximately

ijOO psia^ iiOO^F)^ Ihe loads considered for thLs i l l u s t r a t i o n arei

Eccentric3 t i e ? ;

e^ = .0001 in-

ep « ^0002 in«

e« = .0001 in*

(me)^ and (me)p - balancing weights needed for a close counterbalance of the

above inherent mass e c c e n t r i c i t i e s

F - rotating magnetic unbalance load of 6^0 pounds

(me)-' - thermally induced load of about 3 pounds at 36,000 rpm

(me)„ - localised assembly and thermally induced unbalance of about 3 pounds

at 36,000 rpm

The presented curves show the corresponding onaft ortoits and phase angles

between the total rotor rotating force vector and the maximum displacement

vectors at the journal and capacitance probe locations„ Ourve I represents

the effects of the vectorial load combination as shoi^n in Figure 12^ with

the Curve II representing the response when only the rotating magnetic load

was rotated by 90 degrees from the original positiono Curve III represents

the load condition as Curve I with the load (me)„ removed. This is the load

condition which may be identified with some of the later CRU V units. As it

may be expectedj, the change in localized loads at the flexible rotor section

affects to a significant degree the shaft extension and to a much lesser


7,1.19

degree the c l o s e s t bearing* The i l l u s t r a t e d technicue was used to predict


the coupled rotor-housing responses for a v a r i e t y of load d i s t r i b u t i o n s , balanc-
ing effects and l a t e r for c o r r e l a t i o n s of measured and calculated responses.

The CRU V housing and i t s own suspension system within the t e s t r i g s form an

equivalent mass-spring system upon which the measured housing acceleration

l e v e l s depend. Some of the e a r l i e r CRU housing-housing suspension systems

and the e n t i r e t e s t r i g were s\ib,jected to vibrolator excitations whiich yielded


s a t i s f a c t o r y checks with predicted responses. Later with changing types of

the t e s t r i g s , i t was resorted to a n a l y t i c a l models for predicting the combined

housing-rotor responses*

The obtained c o r r e l a t i o n s between the predicted and measured o r b i t s and accelera-

tions are i l l u s t r a t e d in a single presentation (Figure 13) for several CRU V

u n i t s i d e n t i f i e d as -3D, -10, e t c . An immediate observation can be made from

t h i s presentation that various u n i t s were subjected to different loads and load

distributions. At t h i s time, assuming no problems in instrumentation, which

w i l l be discussed l a t e r , the i t e r a t i o n s witlri s l i f h t l y varied load magnitudes

and d i s t r i b u t i o n s to provide closer correlations between the measured and

predicted a c c e l e r a t i o n s , measured and predicted, shaft extension orbits and

the accompanying phase angle were used to e s t a b l i s h a b e t t e r assesaiaent of

loads. The responses shown in Figure 13 were based on the f i r s t selection of

loads to sliow the agreement of c o r r e l a t i o n s for typical operating conditions.

Operational or t e s t times for various CRU V u n i t s identified in Figure 13 can

be considered as another factor in assessing the obtained correlations of the

o r b i t s and a c c e l e r a t i o n s . These operational times are shown to indicate

the times a t the t e s t completions of the i d e n t i f i e d CRU V u n i t s and should


7.1.20

not be construed as the failing times of the bearings or the system.

Some problems in instrumentation may be suggested by the varying degree of

satisfactory correlations of response ,s of some units. For example, a degree

of poor correlation of the shaft extension orbits in the case of CRU unit ID

related to the ootained satisfactory correlations between the measured and pre-

dicted acceleration levels can be talcen as an inoication of some problems in

capacitance probes and associatca instrumentation. Continuing along a similar

thought, deviations between the predicted and experimental acceleration levels

most probably can be assigned to inaccurately defined or selected load distri-

butions.

At the beginning of the CRU V testi T nroprnm tie desired or required instru-

mentation was not available and some elements had to be improvised and designed,

such as the caDacitance orobe extension stems, probe cooling systems, seals,

etc. VJith time the CRU ¥ test instrums^ntation was progressively improved to

more reliable and effective measuring systems.

Some of the problems experienced in instrumentation related to the measure-

ments of shaft orbits can be assigned to tl e off-site calibration techniques,

shaft runouts, teirporary capacitance probe circuit shortings due to flooding

of the gap, long duration and thermal drifts, etc. Acceleroneters on the other

hand were not affected by most of tie listed factors and correlations between

the measured and calculated resoonses may be considered as reliable indicators

of the bearing reactions.

The directly measured rotor orbits were comprised by tie actual shaft extension

orbit and the physical shaft runout at tie nrobe location. The shaft exten-

sions were ground to less tlan 100 micro-incles (TTH) referred simultaneously
7.1.21

to both journal surfaces. However, in most cases even such a small runout

constituted a considerable portion of the directly measured shaft extension

orbits. Thus, for obtaining a true shaft extension orbit from wl ich the journal

orbits were partly derived, it was mandatory to consider a vectorial relation-

ship between the recorded runout and directly measured orbit at the capacitance

probe locations.

At the beginning of testing it was quite desirable to use a dual monitoring

system such as the accelerometers and capacitance probes to establish confidence

and accuracy levels of the analytical methods and instrumentation in hostile

environment.

In the later CRU Y units utilizing tie established confidence levels in

predicting the coupled rotor-housing resoonses and experience with previous

units, the capacitance probfs were removed leaving the housing accelerometers

as the only monitoring instrumentation of the rotor resoonses.


FIGURE 1 TURBOALTERNATOR ASSEMBLY-CRU V

iP^^
CAPACITANCE
PROBE

-ALTERNATC«
BEARING
^
THRUST
BEARING

TURBINE
BEARING

^
r "
FIGURE 2 CRU V TEST SET-UP

N3
FIGURE 3 BEARING TEST RIG
mKIMIWmBESlWSM MASS mmimi mmm
EAOHSMioNirrmsr BYlNSmLMIOHOFSCRliS

mffiOPTlOML

mmsfiBiMi mmMm mmm mmm mmm


in
^k
'^%c-...,.
•r-'-'-s-.
' >^/' ' / / ' vV'/
' ''///
ONiwaa avd-33aHi p ajnoij
7.1.26

FIGURE 5 TYPICAL COMPONENT LEVEL


BEARING TEST CAPACITANCE PROBE DATA

LUBRICANT : LIQUID MERCURY


LUBRICANT INLET TEMPERATURE : ROOM TEMP.
SUPPLY PRESSURE : 400 PSIA
ROTATING LOAD : .00725 IN-OZ/BRG
UNIDIRECTIONAL LOAD : ZERO
SPEED AS SHOWN

. - 1. _ s \ • » II

;i.-.:.;.
I..

' .^;nv^ •••:•.•••••.[


! -1 5-t .:t'".;,.r...'!.\.'....'i '
*'".'. •••'•*",• i " " • ' " ' ^ ,

•^^^•^•?sms^
1^ ::M^^m^m:^m^:r^

15,000 RPM 25,000 RPM

'l^rfii^v

:\W.\ _»..?«apiv>
VI* ^ti* %

36,000 RPM 45,000 RPM


7.1.27

FIGURE 6 THREE-PAD BEARING FILM STIFFNESS AND DAMPING


COEFFICIENTS (STANDARD BEARING CLEARANCE)

LUBRICAN'l -MERCURY

TEST CONDITIONS:
S U P P L Y P R E S S U R E Z50 AND 400 P S I
ROTATING L O A D S :
.OO^^S AND . 0 0 7 Z S I N - O Z / B R G
T E M P E R A T U R E : ROOM

u
SO
m BEARING F I L M DAxM i n N G
J ^ ^ - -
40
O o ^ ^ r^
Pi
z
< 30

^
Q
20

180

160 /
CO BEARING F I L M S T I l F N E S S
CO
W
Z 140 /

H 120
A[
en

az
(^ en
100
p = 4 0 0 PSI
v#
r ^
rTl^^^
80

60
J^zz^ C ^ '^

p = 250 PSI
40

20
10 20 50 40 50
SHAFT SPEED - KRPM
7.1.28

FIGURE 7 THREE-PAD BEARING FILM STIFFNESS AND DAMPING


COEFFICIENTS (1.3 X STANDARD BEARING CLEARANCE)

1. 3 X STD. BEARING CLEARANCE


LUBRICANT - MERCURY 206 PSIA
TEST CONDITIONS: 256 PSIA
ROTATING LOAD: 406 PSIA
,00725IN-OZ/BRG 456 PSIA
TEMPERATURE- ROOM

50

U 40

30
o
'z Q
« 5 20

0-
160

140
en
W
Z 120
h
H
m 7 100
o
d X
80
o
«
w
pq
40

20

10 20 30 40
SHAFT SPEED - KRPM
7.1,29

FIGURE 8 STABILITY THRESHOLD AT N O LOAD


350

Three-Pod Bearing
No Rotating Load
Lubricant Supply ^ 100°F
300
Drain Pressure- 6 psio
1 .3 X Standard Bearing Clearance

20 30
SHAFT SPEED (RPM X 10 3)
FIGURE 9 STABILITY THRESHOLD WITH UNIDIRECTIONAL LOAD

4 5 6 7 8
BEARING FLOW (LB/MIN)
FIGURE 10 CRU V CRITICAL SPEEDS VERSUS BEARING FILM STIFFNESS CURVES

7
80 ~

70 ~
1 SECOND MODE
CRITICAL SPEED
/
2 /
DS 60 -
^^_^^,^,,*«*'«*'***°FIRST MODE
o / ^t^^"^^^"^ CRITICAL SPEED
V)
D
Z /
<
CO 60 /

U' X
O /

2
^
/
C 40
J
J ^
/
/ / VA ^ y OPERATING SPEED

ii
L«»«««(
a. """^7^ S6, 000 RPM
/
/
<::: 30 , /
/ /
en

I1 1
A 1 / BEARING F I L M S T I F F N E S S CURVES
/ L STANDARD C L E A R A N C E , p= 400 P S I
2. STANDARD C L E A R A N C E , p= 250 P S I
20
3 . L 3 X STANDARD C L E A R A N C E p = 400 P S I

/ i 4 . L 3 X STANDARD C L E A R A N C E , p = 250 P S I

10
!
10-^ 5 6 7 8 9 10' 10° 10'
BEARING F I L M S T I F F N E S S (K). L B / I N
FIGURE I I SHAFT ORBITS DURING TRANSIENT STATE
OPERATION (SPINUPS AND SPINDOWNS)

«_ STANDARD BEARING CLEARANCE

ENVELOPE OF MEASURED
SHAFT EXTENSION ORBITS
FOR VARIOUS CRU V UNITS

C A t C U ^

CALCULATED
JOURNAL «BITS FOR
TYPICAL CRU V LOADS

5000 10,000 15,000 20,000 25,000 30, OX) 35, W )


SHAFT SPEED, RPM
7.1.33

FIGURE 12 TYPICAL LOAD DISTRIBUTION AND


THEORETICAL SHAFT ORBITS AT 36,000 RPM
"'^' i ^
f/^el 1 ('^'•h (mc)^
€,
ffi \
e.3 1

LZZZZH L _ ^ ^
\

——J—-
r
— '
L
H
/ •?•
© © (8
©©

400
0| 191°

1 //

-i
f
300 _ II

200

01 113° ^.--..^^

100 ^ ^ ^ ^ 1 2 8 °
0|,l 140° ^i '''°^^^xJL^
0,1, 163°
© © ®
LOCATIONS LOAD DISTRIBUTIONS
1 ALTERNATOR BEARING (me) I BALANCING PLANE EFFECTS
2 BALANCING PLANE (me)3 BALANCING PLANE EFFECTS
3 BALANCING PLANE Fm ROTATING MAGNETIC LOAD
4 TURBINE WHEEL (me)6 THERMALLY INDUCED UNBALANCE
5 TURBINE WHEEL (me)/ LOCALIZED ASSEMBLY OR
6 TURBINE BEARING THERMALLY INDUCED
7 SHAFT EXTENSION UNBALANCE
8 CAPACITANCE PROBE ei,e2, DISTRIBUTED UNBALANCES
LOCATION
«3
7.1.34

FIGURE 13 COMPARISON CT MEASURED AND PREDICTED


ROTOR RESPONSES AT 36,000 RPM
ACTUAL ORBIT
(MEASURED ORBIT MINUS RUNOUT
AT THE PROBE)
sm ^
PREDICTED
U
- ^ UJ
600

Z
0 12 400
zu
^Z
X
-
I
3D 1C
i
4D
CRU V UNITS
i
2E ID
O
Z

5C

.60
4700 HRS.
2500 HRS,

Z> =40
- y
^Z
go 998 HRS
> < 20 HRS 683 HRS.
760 HRS. ^
.20
eg r-^
u

3D
i
IC 4D
i 1
2E ID 5C
CRU V UNITS MEASURED

^ PREDICTED
7.2.1

CALCULATION AND EXPERIMENTS ON THE UNBALANCE


RESPONSE OF FLEXIBLE ROTORS SUPPORTED BY TILTING
PAD BEARINGS OPERATING IN THE TURBULENT FLOW REGIME

by

J.W. Lund and F.J. Orcutt


Mechanical Technology Incorporated
Latham, New York

ABSTRACT

The results of a combined analytical and experimental investigation of the


unbalance vibrations of a rotor are presented. The analysis applies to a
general rotor-bearing system in which the dynamic bearing forces are repre-
sented by four spring coefficients and four damping coefficients. The rotor
can be represented either as a lumped or a distributed parameter system, and
gyroscopic moments are included. In general, the unbalance whirl motion of
the rotor will be elliptical. The analysis has been programmed for a digital
computer to obtain results for comparison with the experimental data.

The test rotor is a uniform, flexible shaft with heavy wheels mounted
at the ends and in the middle. The rotor is supported in two silicone fluid
lubricated, tilting-pad bearings. The rotor amplitude caused by an induced
unbalance has been measured over a speed range of 3000 to 24,000 RPM for three
different rotor configurations, obtained by removing one or both end wheels.
This speed range extends to or through the third ciritcal speed for each of
the rotor configurations. The results are compared with the theoretical values
and, in general, the agreement is found to be good.

#
INTRODUCTION
To date, the shaft diameters of high-speed, process-fluid lubricated machines
have generally been selected with the intent of maintaining high shaft rigidity,
so that all flexural critical speeds are well above the highest operating speed.
Application of this design approach, however, becomes increasingly difficult
as operating speeds are raised, in addition to introducing needlessly high parasitic
losses in the bearings and seals. Accordingly, it is anticipated that as the
level of experience with high speed, process-fluid lubricated turbomachinery
increases, there will be greater acceptance of the flexible rotor design approach.

Parasitic bearing power losses may be minimized by adopting the flexible rotor
approach to the design of turbomachinery. This is particularly the case with
units that operate at high speeds and are lubricated with low viscosity cycle
fluids, such as water or liquid metals, so that the bearings operate beyond the
transition from laminar flow. In the superlaminar flow regimes, but where the
4
Reynolds Number is less than 10 , the bearing and seal losses vary as the 3.5
power of the diameter. At higher values of Reynolds Number, the exponent is
still larger. Moreover, the use of flexible rotors permits greater machinery
compactness and it may also allow design simplification. For these reasons, it
has long since been generally adopted for high speed, oil lubricated machines
such as steam and gas turbines.

In units that operate at high speeds and where the steady-state load on the radial
bearings is small, reductions in bearing diameter are generally permissible from
the standpoint of load capacity. The effect of increased flexibility of the
rotor will, however, cause highly significant changes in the dynamic response of
the rotor-bearings system. Accordingly, an accurate and detailed investigation
of rotor response to unbalanced loads is necessary during the design phase, in
order to arrive at arrangements that are not subject to high vibration amplitude
within operating speed ranges. Such an analysis must couple the dynamic chara-
cteristics of the bearings and those of the rotor, as well as the pedestals
or other flexible supports.
7.

Analytical procedures for computing the dynamic properties of fluid film bearings
in turbulent as well as in laminar flow regimes are currently available, from
earlier studies of turbulent fluid films 1,2 . These dynamic properties of
fluid film bearings, expressed in the form of stiffness and damping coefficients,
are then coupled with the dynamic properties of the rotor and those of the
pedestals or other flexible supports, to analyze the dynamic response of the
overall rotor-bearings-supports system, using procedures which are also currently
available 3 .

In this presentation, the dynamic response analysis of the general, rotor-bearings


system is described, indicating how the rotor may be treated either by a lumped
or distributed parameter system. In addition, an experimental investigation
that was conducted under AEC Contract AT(30-1)-3363 is described, which was
performed for the purposes of:

(a) evaluting the effectiveness of the combination of theoretical bearing


dynamic properties in turbulait flow regime and the rotor dynamic
response analysis,as a design tool for high-speed rotor-bearing system
design, and

(b) demonstrating the feasibility of operating systems with flexible rotors,


supported on bearings lubricated with low viscosity fluids and operating
in the turbulent flow regime.

The rotor vibration amplitudes caused by induced mechanical unbalances were


measured in five radial planes, over a speed range of 3000 to 24,000 RPM, with
three different rotor configurations. This speed range extended to or through
the third critical speed, for each of the rotor configurations. The results were
in good agreement with the calculated values and it was concluded that the current
existing theoretical analyses of bearing and rotor-dynamics are effective design
tools.

Before proceeding with the review of the analysis and the test data, it is appro-
priate to emphasize here the need for investigating the response of high-speed,
rotor-bearing systems over the entire speed range of interest. It is not generall
sufficient to merely compute the critical speeds, considering the rotor and bear-
ing flexibilities and then adjust the rotor-bearings design so that these critical
speeds lie outside the operating speed range. While critical speed calculations
are ©Wiously valuable, they provide only a part of the information which the
desigaer needs. In particular, the following information is not provided by
critical speed calculations, but may be obtained from calculation of the unbalance
response of the rotor-bearings system, including the effects of bearing damping:

1. There can be significant effects of bearing film damping on the critical


speeds. While bearing flexibility lowers the critical speeds, the
bearing damping usually has the opposite effect. Hence, a conventional
critical speed calculation which includes the bearing flexibility, but
not the bearing damping, will tend to give values for the critical speeds
which are lower than the ones actually observed. The actual critical
speeds can be determined from a series of unbalance response calculations
made at intervals over the speed range to locate the response peak.

2. Depending on the rotor and bearing designs, there will be great diff-
erences in the amplitudes of vibration which are encountered at the
critical speeds. In most cases, bearing film damping is the dominant
controlling Influence on rotor vibration at the critical speeds. Bear-
ing damping can be so effective that, in many machines, it is very diffi-
cult to identify the critical speeds without special instrumentation.
On the other hand, if the bearings ate not effective in controlling the
rotor vibration either because of their location or because they have
insufficient damping, there may be very large amplitudes at the critical
speeds. Thus, unbalance response calculations can be used to determine
the importance of the critical speeds and to evaluate the adequacy of the
proposed rotor and bearing designs.

3. The response of a particular rotor-bearing system will depend on the


amount and distribution of rotor unbalance. By placing the unbalance
at different positions on the rotor when performing unbalance response
calculations, the sensitivity of the rotor to unbalance can be evaluated.
Also, guidelines can be established for specifying the maximum tolerable
unbalance of individual rotor components. In the same way, response cal-
culations may be helpful in choosing the best locations for balancing planes.
7.2.

ANALYSIS *

A typical rotor consists of a shaft on which is mounted a number of wheels,


impellers or similar components. The shaft is seldom uniform but its diameter
changes in steps along the length of the rotor. The rotor is supported in two
or more fluid film journal bearings.

At any given speed and in the absence of rotor mass unbalance, the axis of the
shaft is in a stationary equilibrium with a shape determined by the static de-
flection of the rotor and the eccentricity of the journals in the bearings.
However, when the rotor contains an unbalance, the centrifugal forces cause the
rotor to whirl and deform such that each point of the shaft axis whirls in a
closed orbit around the corresponding equilibrium position. The whirl orbits
are closely elliptical and their size, orientation and phase angle vary along
the length of the rotor. It is the purpose of the analysis to set up a method
for calculating these whirl orbits for an arbitrary rotor subjected to specified
mass unbalances.

In order not to complicate the analysts unnecessarily, the hydrodynamic forces


of the bearing fluid film are linearized. If the journal center amplitudes are
X and y, measured from the static equilibrium position, and the corresponding
dynamic forces are F and F , a first order Taylor series expansion of the
bearing forces can be written:

<i)

The four spring coefficients K ,K ,K and K , and the four damping coefficients
BXX ,Bxy ,Byx and Byy are determined directly
J
from the rpartial derivatives of F^ and
F , evaluated at the journal center equilibrium position corresponding to the
particular rotor speed. They may be obtained by perturbing and solving the lubri-
cation equation (Reynolds Equation) [3,65889,101. Thus, the coefficients depend
on the bearing geometry, the lubricant viscosity, the static bearing reaction and
the rotor speed. They can be expressed In dimensionless form as functions of the

* See Nomenclature on Page 7.2.21.


bearing Sommerfeld number and have been measured and computed for several bearing
types 4 through 12. For most bearing types, the coefficients do not possess
any symmetry characteristics and it is not possible to select a set of principal
axes such that the cross-coupling terms are eliminated [iij. Hence, the whirl
orbit becomes elliptical.

Although Eq. (1), strictly speaking, is valid only for small amplitudes, the
expressions prove to represent the actual fluid film force also for large
amplitudes with surprisingly good accuracy as long as the minimum film thickness
during the whirling motion does not become less than approximately 25 per cent
of the radial clearance 13 I . In an actual application the rotor amplitude is
rarely allowed to be that large and the chosen linearized representation of the
fluid film forces is, therefore, adequate for most practical purposes. This
seems also confirmed by the present experimental investigation as discussed later.

For analysis purposes the rotor itself can be represented by a series of stations
connected by shaft sections of uniform cross-section as indicated in Fig. 1.
The stations are chosen to coincide with any concentrated rotor masses such as
wheels, impellers, etc. and, furthermore, stations are intro<luceci at each location
where the shaft diameter changes, at each bearing location and at both rotor ends.
Thus, each station can be assigned a mass, a transverse mass n»ment of inertia, a
polar mass moment of inertia, an unbalance and a set of 8 bearing spring and damping
coefficients. For any particular station, several or all of these quantities may
be equal to zero depending on which station is considered. Specifically, it should
be noted that the mass of the shaft is treated separately and is not included at
the rotor stations. Then applying a force balance and a moment balance, the jump
in shear force and bending moment across a rotor station can be expressed in terms
of the rotor amplitude and slope at the station. The equations are derived in
Appendix A where they are given as Eqs. (9) to (12).

For each shaft section the length, the cross-sectional area, the cross-sectional
moment of Inertia and the shape factor for shear deformation are known. The shaft
material properties: Youngs modulus, shear modulus and mass density, are also
given. Then, the relationships for the amplitude, the slope, the bending moment
and the shear force between the two ends of the shaft section can be established
7.2.

as shown by Eqs. (40) to (43) in Appendix B. The coefficients in these equations


are speed dependent and include th@ effect of shaft inertia and shear deformation.

The equations for the rotor stations together with the equations for the shaft
sections establish a set of recurrence formulas by which a step-by-step calcula-
tion of the rotor can be performed. Assuming the rotor ends to be free, a succ-
essive application of the recurrence formulas allows coaaputing the amplitude, the
slope, the bending moment and the shear force along the rotor in terms of the
amplitude and the slope at the first station. Adding the contribution from the
unbalance forces and equating the bending moment and the shear force at the last
station to sero, makes it possible to determine the unknown amplitude and slope at
the first station. By back-substitution, the amplitude at each station can finally
be calculated. The procedure is given in more detail in Appendix A and is readily
progranmed for a digital computer. Such a program has been used to obtain the re-
sults for comparison with the experimental data as discussed later.

The analysis is seen to be an extension of the Prohl Method 14 for calculating


the critical speeds of a rotor. The present method differs from the original one
by including the anisotropic stiffness and damping characteristics of the fluid
film bearings whereby the rotor motions in the horizontal and vertical directions
become coupled and the whirl orbits become elliptical rather than circular. This
feature also necessitates a different treatment of the gyroscopic moments as shown
in Appendix A. Furthermore, in the original method the mass of the shaft is
lumped at the rotor stations and the shaft is represented by its static Influence
coefficients, whereas the present method includes the effect of the distributed
shaft mass and employs dynamic influence coefficients. Thus, within the assumptions
Inherent in conventional beam theory, the present method gives an exact solution
to the problem where the original method involves approximations. It must be noted,
however, that the more accurate treatment is only necessary if the number of rotor
stations is less than approximately four to five times the number of that critical
speed which follows after the upper limit of the operating speed range of the rotor.
In this connection it may be observed that the solution by the present method, and
also by the Prohl Method, depends directly on evaluating a determinant which is
essentially the characteristic determinant of the system. For illustration, con-
sider a simply supported, uniform shaft of length i. The characteristic deter-
minant as computed from the analysis is:
8

where B is proportional to the square root of the speed and is defined by


Eq. (31) in Appendix B. The determinant is zero for (fl|) -NITwhere m then
gives the number of the critical speed. The characteristic determinants for
other cases are analogous. It is seen, however, that the actual value of the
determinant is the remainder of a subtraction involving:

Thus, if the computer works to an accuracy of k significant figures the re-


mainder will be lost when 4" S I'ft h f|3 f) - 10 or when i^l}-^ 2+2,3'k ^ i.e.
when 1rf\^ fto + 0,7°k. Most computers carry eight significant figures and
in that case the present method suffers an Increasing loss of accuracy beyond
the fourth critical speed until it finally breaks down completely around the
sixth critical speed. Going to double precision ( k-16 ) raises the ultimate
limit to around the eleventh critical speed.
I
For the calculations of the present test rotor, these considerations are of little
Importance. Furthermore, the test bearings are tilting pad bearings with four
shoes arranged symmetrical with respect to both a vertical line and a horizontal
line. Thereby the spring and damping coefficients become the same in all
directions and the rotor whirl orbits are circular. This greatly simplifies
the analysis and the calculations.
7.2

EXPERIMENTAL APPABATUS

The test rotor is a cylindrical steel shaft with overall length of 41 inches and a
diameter of 2.5 inches except for a centrally located integral 4isk of 6 inches
diameter and 6 inches length. Detachable disks of 6 inches diameter, 3 inches
length can be mounted on the ends of the rotor by heavy interference fit and lock
nuts which clamp them against shoulders on the shaft. The total weight of the
rotor, exclusive of the detachable end masses, is 88 lbs. The center disk, not
counting the inner 2,5 inch diameter section, weights 36 lbs. and the end disks
each weigh 18 lbs.

The rotor is supported by two tilting-pad journal bearings whose centerlines are
12.5 inches on either side of the rotor center plane. These are four-pad bearings
with slenderness ratio (L/D) of 1.0 and bearing diameter of 2.5 inches. Other
design characteristics include:
-3
Clearance ratio (C/R) - 3 x 10 in/in (based on machined clearance)
Pad arc length - 80 degrees
Pivot position - 0.55 of the arc length (44 degrees) n^asured from the
leading edge.
Geometrical preload coefficient - 0.5

Geometrical preloading is used to give improved bearing film stiffness at high


speeds with low steady-state loads. It is achieved by moving the pads radially
inward so that the actual clearance at the pivot location is less than the
machined clearance (based on pad and shaft radii of curvature). In this case,
the clearance at the pivots is just half the machined clearance.

The pivot configuration is a sphere seated in an internal cylindrical surface


machined into the back surface of the pad. The bearings are oriented in the
housing so the steady-state, gravity, load line bisects the pivot positions.
That is, the pivots are at 45 degrees from a vertical line drawn through the
bearing center.

Axial positioning of the shaft is accomplished by externally-pressurized, gas-


lubricated thrust bearings on either side of the center disk.
/.2.10

The bearings are rigidly mounted in housings machined from a single block of
aluminum jig plate. The housings are bolted and keyed to a massive structural f|
4
steel base whose moment of inertia about the horizontal axis exceeds 9,000 in .
There were no detectable traces of flexure or vibration of the base during the
experiments.

The rotor is driven by an electric motor through a crowned spline coupling. The
coupling can accommodate misalignment, between the motor and rotor axes, of up to
0.030 inch without restraint to the shaft. The motor is bolted and keyed to the
same base surface which supports the bearing housings. Motor and test rotor
shafts were aligned with less than 0.002 inch total runout measured radially and
0.0015 inch total runout measured axially on a 2 inch radius. A variable-frequency
motor-generator set supplying power to the motor provides for variable speed
operation within a range from about 3000 to 24,000 rpm.

The journal bearings are lubricated by a very low viscosity silicone fluid, 0.65
cs at 77F and 0,51 cs at 130F. Lubricant at controlled temperature is pumped into
the bearing housing so the bearings are operated submerged in the lubricant. There
are clearance seals and scavenging rings on either side of the bearing housings.

The motion of the rotor is measured in both horizontal and vertical directions at
each of five measurement planes along its length. There are measurements planes
at each end, at the center, and at the inboard sealing ring of each bearing housing
(8.65 in, from the rotor center plane). These measurements are obtained with eddy-
current, non-contacting proximity sensors. Readout is by oscilloscope with the
signals connected to the horizontal and vertical deflection plates of the cathode
ray tube. With this arrangement, the rotor center axis at the measurement plane is
represented by a spot on the oscilloscope screen and motions of the rotor axis are
reproduced by the spot on the screen.

Accurate measurement of the response of the rotor to imposed unbalance load requires
care in minimizing other causes of real or apparent shaft axis motion. The principal
such causes are shaft surface out-of-roundness or non-concentricity and residual
rotor unbalance. Rigid specifications were imposed on rotor surface roundness and
concentricity with the result that subsequent gaging showed all surfaces to be
7.2

round and concentric with the bearing journals within less than 0,00012 inch total
runout. The rotor was balanced first in a balancing machine and then in place in
the apparatus. In-place balancing was done by a trial and error procedure. The
single, centered, disk rotor was balanced so that the maximum orbit amplitude at
any measurement plane over the entire speed.range was 0.00014 inch radius. For
the two-disk system, which is more sensitive to unbalance, the maximum orbit
radius was 0.00021 inch at the light end and all other positions were always below
0.00015 inch radius. The three-disk system was balanced at speeds up to 10,800
rpm. At this speed, the maximum whirl orbit radius was 0.00035 inch. At all
speeds below 10,000 rpm, the maximiun orbit radius was below 0.00013 inch.

In order to minimize the effects of runouts and residual unbalance on the test
results, it is desirable to Impose large deliberate unbalances to obtain sizeable
response orbits. On the other hand, the assumption of linear bearing stiffness
and damping about the steady-state rotor position will be strained if the orbits
are too large. For most operating conditions, a good balance between these con-
siderations was considered to be: the larger of the orbit radii at the two
stations adjacent to the bearings should be between 0.20 and 0.40 times the bearing
radial clearance measured at the pivots, or, between about 0.0004 and 0.0008 Inch
orbit radius. This limit was not applied to the three-disk rotor for reasons which
will become clear. During the experiments, the amount of unbalance weight was
adjusted as the speed was changed In order to remain within the limits given above.
Since the orbit amplitudes at the center and end positions were usually larger than
those at the bearing positions, the orbits with Imposed unbalances were usually at
least 4 times larger than the corresponding orbits with no Imposed unbalance.

BeatMjig temperatures were measured by thermocouples welded to the trailing edge of


e a A ©f two pads,flush with the bearing surface. These measured bearing temper-
at^t'es were used to establish the lubricant film viscosity which, In turn,Is needed
to specify bearing stiffness and damping from the bearing design data* Bearing
temperatures ranged up to about 130F at full speed in the experiments. Within
limits, bearing temperature could be controlled by varying lubricant flow and in-
let temperature.
CALCULATION OF TEST ROTOR UNBALANCE RESPONSE

A computer program based on the preceeding analysis was used to calculate the
unbalance response of each of the three rotor configurations, that Is, with one,
two or three disks. Dynamic stiffness and damping values for the test bearings
were obtained from the data given in Ref. 5- The stiffness and damping vary with
the shaft speed and steady-state load and, since the test bearings operate in the
turbulent flow regime, with the mean film Reynolds Number. In the case of the
four-pad bearing with steady-state load line midway between the pivots, synmetry
results in elimination of the cross-coupling terms and equality of stiffness and
damping in x and y directions. Sample dimensional stiffness and damping values
are given in Table 1 to illustrate the values obtained.

The computer program output included the calculated vibration amplitude at each
of the planes where experimental measurements were made. Because of test bearing
symmetry, the calculated vibration is a circular orbit whose radius is given In
mlls/oz-in. total unbalance.

Flexibility of the pad pivots was Introduced into the calculations as a stiffness
in series with the bearing. The calculated Hertzian contact stiffness of the
combined pivots at 100 lb. bearing loa<
load is 5 x 10 lb/in. and this is the value
used in the calculations at all speeds,

The undamped critical speeds of the three rotor-bearing systems were calculated
using the Prohl method (Ref. 11) together with the theoretical bearing stiffness
data. This was done for comparison with the calculated and measured unbalance
response data. The comparisons are significant since it is common practice to
use calculated critical speeds in rotor-bearing design analysisjbut much less
common to perform unbalance response calculations. The results are tabulated
below for each of the three rotor configurations.

Rotor Configurations Calculated Critical Speeds by Prohl Method


1 2 3
7200 17,800 24,500
One-Disk Rotor
6100 8,700 21,000
Two-Disk Rotor
5000 5,800 11,000
Three-Disk Rotor
TABLE 1

THEORETICAL TEST BEARING STIFFNESS AND DAMPING

Stiffness, lb/in.
(1) One-Disk Two-Disk Rotor Three-Disk
Speed,rpm Rotor Light-End Heavy-End Rotor

6050 8.1x10^ 8.1x10^^ 8.4x10^ 8.5x10*


8100 10.6 10.8 11.8 12.3
11,000 15.3 15.4 15.6 15,8
16,000 26.2 26.8 27.9 29.0
24,000 50.3 50.6 51.2 51.4

Damping, lb.sec/in.

6050 148 156 168 170


8100 156 161 170 174
11,000 173 178 183 186
16,000 218 223 228 231
24,000 266 269 273 274

(1) The bearing film viscosity Is assumed to be 0.65 cs at speeds up to


9000 rpm, 0,60 cs from 9000 to 17,000 rpm and 0.53 cs up to 24,000
rpm. Bearing temperature was controlled during the experiments to
match these values.
CCTiPARisoN 0¥,,EMM£^mM^.mQ.MiMIMm.MmA.M.^mmjm..MMLm^ 1
Calculated curves and experimental data points of rotor response as a function
of speed for the single-disk rotor are given in Figs. 2 through 4. The location
and relative magnitudes of the imposed unbalance weights are indicated schemat-
ically on the figures. Because of symmetry, the response at corresponding
measurement planes on either side of the centerline should be the same and so the
results at both ends are given together. At each measurement plane there are two
sets of data. These are from two separate series of runs in which two different
angular positions of unbalance weights were used. That is, after the first series
was obtained, all of the unbalance weights were i»ved to a different angular
position on the rotor, usually opposite the first location. Differences between
these sets of data points can be taken as an indication of the influence of runout
and residual unbalance effects on the experimental results.

The theoretical curves show the one-disk rotor system to be well behaved through
the operating speed range. The orbit amplitudes level off about midway between
the first and second calculated critical speeds and remain about the same or
gradually decline through the remainder of the operating range. The experimental
data show the same features and the absolute levels of vibration amplitude agree
rather well with the calculated levels. While there is some experimental scatter,
the measurements show no marked differences between the coupling and free ends
or between the two angular positions of the unbalance weights.

According to theory, the response orbits should always be circular. The actual
measured orbits were not always so, although their ellipticity was seldom marked.
There were no recorded instances when the ratio of major to minor axes was greater
than 3 and very few when it exceeded 2. When the orbits were elliptical, the major
and minor axes were measured and the mean between them was taken as the experimental
data point. There are a number of possible reasons for non-circular orbits of which
the most significant is believed to be asymmetry of the bearings, because of
slightly different preload between the two opposing pairs of pads.

Calculated curves and experimental data for the two-disk rotor with the unbalance
weights in-line are given in Figs. 5 through 9. This configuration experiences
7.2.15

significantly higher vibration amplitudes for an equivalent total unbalance weight


than the one-disk rotor. Also, there are well defined vibration peaks of a
translation whirl mode at about 10,000 rpm and a conical whirl mode at about
12,000 rpm. The measured behavior of the rotor is consistent with the calculated
results and, with one exception, the measured and calculated vibration amplitudes
agree quite well. At the light end, one of the two sets of data shows orbits
which are substantially larger than either the calculated results or those for
the other set of data. There was no obvious explanation for this.

Results for the two-disk rotor when the unbalance weights at the two disks are 180
degrees out-of-line are given in Figs. 10 through 14. This results in a sharply
defined conical mode vibration peak at about 11,000 rpm with noticeably larger
amplitude than occurred when the unbalance weights were in-line. Again, agree-
ment between experimental and theoretical results is good with respect to both
the trend and the whirl orbit amplitudes.

At the third, free-free mode, critical speed of the three-disks rotor system, the
bearings are very close to the rotor nodal points. The result is a sharp peak
with very large whirl orbit amplitudes at the calculated third critical speed.
Whirl orbit amplitudes for the out-of-line arrangement of unbalance weights are
shown in Figs. 15 through 17.

The total unbalance weight used when passing through the response peak was about
0.18 in-oz which results in very small orbits and comparatively poor agreement at
higher speeds when the response amplitudes are much lower. When passing through
the third critical speed with this small unbalance, the or-bits at the stations
adjacent to the bearings were appreciably larger than the bearing clearance.
Because of concern over the well-being of the apparatus and its operators, the
rotor was accelerated through the response peak and there was no steady-state
operation at that point. Peak amplitude data were obtained by using a camera
attachment for the oscilloscope with the shutter held open during the passage
through the peak. The largest recorded orbit amplitude is shown as the data
point at the peak in Figs. 15 through 17,

The response of the three-disk rotorwith the unbalance weights in-line is


similar in trend to the results when the weights are out-of-line. The amplitudes
are considerably lower though they are still very large. Agreement between cal-
culated results is good for both cases and there was precise agreement on the
speed at which the response peak occurred.

The modes of rotor vibration during the experiments were determined from obser-
vations of the phase angles between the direction of the unbalance force and the
response at the different measurement planes. The measurement was accomplished
by using a sensor which responds with a sharp voltage peak as a mark on the shaft
passes. The amplified voltage peak is used to momentarily dim the beam of the
oscilloscope so that there is a short interruption of the orbit on the screen
which identifies the location of the rotor axis on its orbit at the time the mark
on the shaft passes the sensor. The phase angle is measured on the oscilloscope
as the angle between the break in the orbit and the known angular position of the
unbalance weights at the time the mark passes the sensor. The accuracy of the
measurement is adversely affected if the orbit is not truly circular. For this
reason and because calculated phase angles are of comparatively little significance
in design analysis, the emphasis has been placed on the response amplitude results.
However, a typical sample of the agreement between measured and calculated phase
angles is given in Table 2 for the two-disk rotor with the unbalance weights
in line.
I
TABLE 2

MEASURED AND CALCULATED PHASE ANGLES


FOR TWO-DISK SYSTEM. UNBALANCES IN-LINE

Calculated

Speed, rpm Light End Light End Center Heavy End Heavy End
Bearing Bearing
3000 30 32 32 34 195
6000 80 89 90 90 220
9000 124 136 146 170 280
10,000 131 146 158 195 288
11,000 137 154 163 260 318
12,000 143 158 163 340 340
15,000 156 168 158 126 33
18,000 163 174 164 134 60

Measured

3000 40 310 20 30 200


6000 63 65 65 58 175
9000 65 80 120 123 200
10,000 100 95 107 100 240
11,000 110 113 168 220 300
12,000 145 100 160 355 30
15,000 168 160 135 132 35
18,000 165 185 160 160 110


DISCUSSION

Comparison of the measured response of the test rotor-bearing system with the re-
sults calculated from the combination of the rotor response analysis described in
this paper and theoretical bearing stiffness and damping has shown good agreement.
In particular, there is agreement in the following respects

(1) For each of three rotor configurations, the trend of the response to
unbalance as the speed was varied over a range extending to or through
the third undamped critical speed was substantially as calculated,

(2) With few exceptions, the measured and calculated amplitudes of vibration
are comparable. Examination of either measured or calculated results for
the purpose of determining the relative sensitivity to unbalance of the
three different rotor configurations clearly leads to the same conclusions.

(3) One application of rotor response calculations Is to determine the sensi-


tivity of the rotor to unbalance at different locations. Substantial
differences in the response of the experimental rotor were predicted
depending on whether the unbalance weights were in-line or out-of-line.
The experiments showed differences which were in good agreement with
these theoretical predictions.

Comparison of unbalance response characteristics with the calculated, undamped


critical speeds is interesting, especially since critical speed calculations are
usually the only type of rotor dynamics analysis performed in design. The three-
disk rotor results show the large amplitude vibration peak occurlng precisely
at the calculated third critical speed. At this vibration mode, the bearing pro-
perties have little effect on the rotor vibration because the bearings are very
close to the nodal points. For this reason, the agreement between calculated
critical speed and rotor response in this Instance is principally a verification
of the accuracy of the rotor response program in calculating the characteristics
of the rotor. When the bearings exert considerable control over the rotor vibra-
tion, as with the one and two disk rotors, the response results,both calculated
and measured, show much less correlation with critical speed calculations. The
peak response with the one-disk system occurs about midway between the first two
7,2.19

critical speeds. Results for the two-disk system with unbalance weights in-line
are especially Interesting since they show two well defined peaks whose mode
shapes correspond to those of the first two critical speeds. These vibration
peaks occur at 1.7 and 1.4 times the first and second critical speeds respectively.
These differences between critical speed and rotor response results are caused
by the effects of bearing damping which is not included in the critical speed
analysis.

A principal assumption in the theoretical analysis is that the bearing stiffness


and damping are linearized about the steady-state rotor position. This assumption
was examined during the experiments by operating with different values of unbalance
weight under otherwise identical conditions. Proportionality between unbalance
weight and response held up within 20 percent or less for weight changes of as
much as 3 to 1 giving orbits as large as 0.6 times the bearing clearance.

These results are regarded as an encouraging demonstration of the feasibility of


using the rotor response analysis and theoretical bearing properties for design
analysis of proposed high-speed rotor-bearing systems. Moreover, they have shown
that rotor response calculations are a valuable supplement to critical speed cal-
culations. Undamped critical speed calculations may not accurately identify
the speeds at which maximum vibration amplitudes will occur and they do not give
any indication of the relative amplitudes of vibration at the different critical
speeds or between different rotor configurations. Rotor response calculations
provide a direct indication of the speeds at which peak vibration amplitudes
will occur and they can be used for comparison of alternate rotor-bearing con-
figurations and for determination of the relative effects of unbalance at
different locations on the rotor.
ACKNOWLEDGEMENTS

The experimental portions of this paper are taken from research supported by
the Atomic Energy Commission under Contract Number AT(30-1)-3363. The
authors wish to thank Mr. N. Grossman of the AEC and Mr. E.B. Arwas of M.T.I,
for their support and guidance in carrying out the experimental work.
7.2.21

NOMENCLATURE

A Cross-sectional area of shaft section, in

CI, to d.} Influence coefficients, see Eqs. (44) to (50)

(Xf b Major and minor semiaxes of elliptical whirl orbit, inch

•^^y J''yw ^ "'*y y Dxx Bearing damping coefficients, lbs.sec/in.


2
E Youngs modulus for shaft section, lbs/in .

M Shear modulus for shaft section, lbs/in .


T 4
1 Cross^ectional transverse moment of inertia of shaft section,in .

ip Polar mass moment of inertia at rotor station, lbs.in.sec .

i"r Transverse mass moment of inertia at rotor station, lbs.in.sec .

Jp Polar mass moment of inertia per unit length of shaft


section, lbs.sec .
jj Transverse mass moment of inertia per unit length of shaft
section, Ibs.sec^.

"^xxy'^Xi<>'ti)X;Kifa Bearing spring coefficients, lbs/In.

k;^«/^3 Static influence coefficients, see Eqs. (51) to (53)

Length of shaft section, Inch

M Bending moment, lbs.in.


2
m Mass at rotor station, lbs.sec /in.

•1 Index giving the rotor station number

fl Index for last rotor station


t Time, sec.

t* =Wr"*"lU/v . Mass unbalance at rotor station, lbs.sec

Wy^Wi^ Components of unbalance along directions fixed in rotor,lbs.sec^


2

I* U Mass unbalance of shaft section per unit length, lbs,sec /in.


1.1.11

I
Nomenclature (cont'd) •I
V Shear force, lbs.

YM Rotor amplitudes, inch

Z Axial coordinate, inch

Zxy^Zy,.^Zj.x;ZLu Zxy-ICxy + rCoBjfV.^ etc. Bearing impedance, lbs/in

c^ Cross-sectional shape factor for shear deformation


•1
See Eq. ( 3 1 ) , inch

j3, ^ ^ 2 See E q s . (34) and ( 3 5 ) , inch"

y Angle between x-axis and major semiaxis of elliptical orbit

cf See Eq . (32) , inch

© x-component of rotor slope, radians

X, ^Aa See Eq. (54)

jX Inertia of shaft section, Eq. (55) , lbs/in.

o Mass density of shaft material, lbs.sec /in

y-component of rotor slope, radians

'^ Phase angle for whirl orbit, see E q s . (19) and (20)

Co Angular speed of rotor, radians/sec.

Indices

Y\ Refers to rotor station n or to shaft section between

station n and (n+1)

X x-component

\A y-component

C Real part (WS (wt'-component)

5 Imaginary part (negative Sin(tot)-component)

•I
7.2.23 \

I
Nomenclature (cont'd)

Superscripts

Refers to bending moment or shear force, just to the right


of a rotor station, see Fig. 1.

Used In few places to denote complex quantity. Note, that


the bar notation is implied but dropped in many equations,
notably Eqs. (9) to (15).

••
REFERENCES

1. E.B. Arwas, "Topical Report on Turbulent Lubrication," Topical Report under


A.I. Subcontract N2-S9-1599 of AEC Contract AT(Il-l) GEN 8, MTI Report 64TR67,
November 1964,

2. H.G. Elrod, Jr., C.W.Ng and C.H.T. Pan, "A Theory of Turbulent Films and Its
Application to Bearings," Topical Report under AEC Contract AT(30-I)-3363,
AEC Report No. NYO-3363-2, MTI Report 65TR9, March 1965 (Also published
under the same title as ASME Paper 66Lubl2, June 1966).

3. J.W. Lund et al: "Rotor Bearing Dynamics Technology," Final Reports under
USAF Contract No. AF 33(615)-1895, U.S. Air Force Reports AFAPL-TR-65-45
Parts III and V, MTI Report 65TR14 and 65TR15, May 1965.

4. F.K. Orcutt and E.B. Arwas, "Analysis of Turbulent Lubrication, Volume I -


The State and Dynamic Properties of Journal Bearings in Laminar and Turbulent
Regimes," 1st Volume of Final Report under NASA Contract NAS--W-771, MTI Report
64TR19, May 1964. (Condensed version published as ASME paper 66-LUBS-4,
"The steady state and dynamic characteristics of a full circular and a partial
arc bearing in the laminar and turbulent flow regimes, June 1966).

5. (a) F.K. Orcutt, "The Steady State and Dynamic Properties of the Tilting Pad
Bearing in the Laminar and Turbulent Flow Regimes," Topical Report under
NASA Contract NAS-w-1021, MTI Report 65TR32, June 1965 (condensed version
published under the same title, as ASME Paper 66 Lub 19, June 1966).

(b) F.K. Orcutt and C.W. Ng, "Steady State and Dynamic Properties of the
Floating Ring Bearing in the Laminar and Turbulent Flow Regimes,"
Topical Report under NASA Contract NAS-w-1021, MTI Report 65TR33,
June 1965,
6. B. Sternlicht, "Elastic and Damping Properties of Cylindrical Journal
Bearings," JOURNAL OF BASIC ENGINEERING, TRANS. ASME, Series D, Vol. 81,
1959, pp. 101-108.

7. A.C. Hagg and G.O. Sankey, "Some Dynamic Properties of Oil-Film Journal
Bearings with Reference to the Unbalance Vibration of Rotors," JOURNAL
OF APPLIED MECHANICS, Vol. 23, TRANS. ASME. Vol. 78, 1956, pp. 302-306.

8. A.C. Hagg and G.O. Sankey, "Elastic and Damping Properties of Oil-Film
Journal Bearings for Application to Unbalance Vibration Calculations,"
JOURNAL OF APPLIED MECHANICS, Vol. 25, TRANS. ASME. Vol, 80, 1958, p. 141.

9. P.C.Warner, "Static and Dynamic Properties of Partial Journal Bearings,"


JOURNAL OF BASIC ENGINEERING, TRANS. ASME, Series D, Vol. 85, 1963,pp.247-257.

10. P.C. Warner and R.J. Thoman, "The Effect of the 150-Degree Partial Bearing M
on Rotor-Unbalance Vibration," JOURNAL OF BASIC ENGINEERING, TRANS. ASME, ^
Series, D, Vol. 86, 1964, pp. 337-347.
11. J.W. Lund and B. Sternlicht, "Rotor-Bearing Dynamics with Emphasis on
Attenuation," JOURNAL OF BASIC ENGINEERING, TRANS. ASME, Vol. 84, Series D,
1962, pp. 491-502.

12. J.W. Lund, "Spring and Damping Coefficients for the Tilting Pad Journal
Bearing," ASLE Transactions, Vol. 7, 1964, pp. 342-352.

13. J.W. Lund, "Self-Excited, Stationary Whirl Orbits of a Journal in a Sleeve


Bearing," PhD Thesis, Rensselaer Polytechnic Institute, Department of
Mechanics, Troy, N.Y. 1966.

14. M.A. Prohl, "A General Method for Calculating Critical Speeds of Flexible
Rotors," JOURNAL OF APPLIED MECHANICS, Vol. 67, 1945, pp. A-142 to A-148.
7.2.26

APPENDIX A

This appendix describes the analysis for calculating the whirl amplitudes of an
unbalanced rotor. The analysis takes into account the anisotropic stiffness
and damping characteristics of the bearings which cause the rotor whirl orbit to
be elliptical. Furthermore, the gyroscopic moments of the rotor wheels are in-
cluded. The analysis is an extension of the Prohl Method 14 and is readily
programmed for a digital computer.

To define the rotor motion, introduce a cartesian coordinate system with the
X-axis vertical downwards, the y-axis horizontal and the z-axis along the rotor
axis. The origin of the x-y-system coincides at each axial location with the
static deflection of the rotor, corrected at any given speed for the eccentricity
of the journals in the bearings. Thus, the rotor amplitudes are x and y, the
corresponding slope components are 6 and w , and the bending moment and the
shear force in the rotor are denoted as M and V, respectively. They have the
components M and M , and V and V .
X y X y

The rotor is represented as a series of stations connected by shaft sections of


uniform cross section. Each station is assigned a mass lii„ , a transverse mass
moment of inertia Ij^ ^ a polar mass moment of Intertla Ipn g a bearing reaction
(expressed by Eq. (1) ) and an unbalance force. The latter force can be deter-
mined by Introducing two mutually perpendicular, fixed directions in the rotor,
denoted as the |-axis and the '^-axis. The instantaneous angle between the
I" -axis and the X-axis is Cot, where 05 is the angular speed of the rotor and t
denotes time. The rotor unbalance can then be defined by its components Mr and
L(^ _, and the corresponding forces, measured in the x-y-system, become:

(3)

CO Utf = U? U^ sin (wt) + (J- U^ C09 (cot)

Considering the n'th rotor station as shown in Fig. 1, a force equation and a
moment equation yields:
IT» 5^+<"!,. f ' Mj,-M,„ (6)

The rotor Is forced to vibrate by the unbalance forces with the frequency CO.
The vibration Is harmonic whereby the amplitude, the slope, the bending moment,
the shear force and the unbalance can be expressed In complex notation:

where: vC-U|+t U ^ ^ X sXc4t>f| j^'tc. ^ and only the real parts apply.

Similar definitions hold for the y-components. Further, define:

Zxyf-Kx'^i'^^)'^ (8)
and analogously for Z)fjj ^Zfejf and Zjj^ , With these definitions, and dropping the
bar notation, Eqs. (4) to (7) become:
.28

V^l ^ V^,, - Z^xi. y« +(toX-Z^^ J ^ h "^^^^K (10)

H»* M,», -W^IT^ e^ + ifci*Ip», f^ (11)

Hj«- M^«-iVIp, ©^ - w* Ir^ f^ (12)


These equations express the jump in shear force and bending moment across a
rotor station.

The n'th station is connected with the (n+1)'th station by a shaft section of
length- C|^ with a cross-sectional area An , a cross-sectional moment of Inertia
Ih and a cross-section shape factor rt^ for shear deformation. The shaft
material has a mass density ^^ , Youngs modulus C.^, and shear modulus Q^ « Then,
as shown in Appendix B, it is possible to express ^niii^^+i j H ^ i|.^Eid Vxin-^l
as linear combinations of X^fO^^ ^ tiHn ^^^ ^ (the relationships for the
y-components are identical). The equations are given as Eqs. (40) to (43) In
Appendix B. The coefficients in the equations are functions of the shaft pro-
perties and the speed of the rotor. For simplicity, It is assumed that unbalance^
rotary Inertia and gyroscopic moments in the shaft can be Ignored, and that
these effects instead are included at the rotor stations.

Equations (9) to (12) together with Eqs. (40) to (43) are a set of recurrence re-
lationships from which the rotor amplitudes can be computed. Assuming the rotor
to have free ends, the bending moments and shear forces at the ends are zero:

(13)

where station O is the last rotor station. Selecting ^i^HijO^ and ^j as


unknowns, repeated application of Eqs. (9) to (12) and Eqs. (40) to (43) results
in:
7.2.29

1;I-W:H: (14)

M)r|'
n
= {^H!:+ =0 (15)

^nJ
where {<^ls a 2 by 4 matrix and d j l s a 4 by 4 matrix. The matrix elements C^l,
and dy are complex. They are found simplest by performing a total of 5 rotor
calculations. In the first calculation, set Vj-sj and Mj - @i = 0^} - Uj, - (?
whereby C^n X^^i ^du^dti ^ d^j and cf^j are obtained. Next, set ^j-' and
y^"^&ir ^j~ U.^~Q and determine the second columns of the matrices. Proceeding
In this way, all the coefficients are obtained. Solving Eq. (15) for ^ij^ij®i
and (Ij allows computing Xj, and u^ from Eq. (14) for all the rotor stations.
Noting that:

XH- CHK.-^iK
"Sw

y^ - K ^'*^*~ KH <^^^(^^^ -Xst, Sin (tat)

and similarly for U^ the semiaxes and the orientation of the ellptlcal whirl
orbit are calculated from

*/g

(17)

(18)

(19)
30

where a^ Is the major semlaxls, b^ the minor semlaxls, If^ the angle from the
x-axls to the major semlaxls In the direction of rotor rotation, and '^ is the
phase angle. The definition of the phase angle is such that If the x-y-coordinate
system Is rotated the angle >!, into an x'-y'-system (i.e. x' is along the major
semlaxls), then the rotor motion can be expressed as:

(20)

If the value for the minor semlaxls Is negative, the rotor precesses backwards.

For those bearing types where £.^y^~ c.u^ and ^y«"'"^«fy It Is seen from
Eqs. (9) to (12) that:

\j~-Ly (f--ie (2i)

Hence, the whirl orbit becomes circular and the outlined calculation procedure
can be considerably simplified. This condition applies to the four shoe tilting
pad bearing used In the present tests.
7.2.31

APPENDIX B

In this appendix a uniform shaft section Is considered and the relationships are

established for the amplitude, slope, bending moment and shear force at one end

of the section In terms of the corresponding equantitles at the other end-

Including shear deformation, the rotation @ of a shaft element becomes:

(22)
® " o ^ ^ "^ar
and similarly for the y-component, where Q Is the modulus of shear and «4 Is
a cross-sectional shape factor ( d ^ 0.75 for a circular cross section).
The bending equation Is given by:

H.^ElIf (23)

and s i m i l a r l y for the y-component. When the shaft Is subjected to unbalance


forces per unit length of WU,, and w W a force equation yields:

jA | p = - 1^ +WW (24)

and analogously for the y-dlrectlon. Finally, if per unit length the shaft has
a transverse mass moment of Inertia |-- and a polar mass moment of Inertia la
a moment equation results In:

Assuming the shaft section to have co istant cross-sectional properties,


Oj^ jM and V can be eliminated from the equations. In this way the
equations governing the shaft motion become:
.32

t A az^ T-^ UQ ^ j i i ^^yf + ?^ af ^ dG it^ ^^r dm ^ dS aF

= c/fu,+ . » ^ i k + i l k i i ^ ^ l l i k ] (27)

FTi\^«T/4+M A + . A A + ^ ^ +411 -^»»^£ip4

- ^ l ^ + ^GA at^ ^ a at dQA at^J ^^^^


Since the analysis is restricted to forced vibrations with a frequency equal
to the angular speed of the shaft,CO , a complex notation can be used:

and similarly for the slope, the bending moment and the shear force. Only the
real part applies. Furthermore, setting:

results in .* U.^ =• -iM.e**• , Thus^quations (27) and (28) can be written:


Although these equations may be solved, the resulting solution Is Impractical.
Instead, It shall be assumed that the shaft section is free of unbalance forces
\ U = O I , and that the effect of rotary Inertia and gyroscopic moment can be
neglected (JT'IP'^ ) • 1^ these effects are significant they can be accounted
for with good accuracy by lumping them at the ends of the shaft section (i.e.
at the rotor stations). It should be noted that even though It is simple to
keep the rotary inertia terms In the above equations, this Is not permissible
without also Including the gyroscopic moments since they are of the same magnitude.

With these assumptions, Eqs. (29) and (30) become Identical and It is only
necessary to consider the first equation. The following definitions are Introduced:

^ = (If^f
/= gl
(31)

(32)

Dropping the bar notation and setting *A-J.J-* jp-0^ Eq(29) becomes:

The characteristic equation has the roots: i ^ , and "ilR^ where:

^.^^lihUf^f -^P^f (34)

/S, = |3[\/l^(J^ -^Ufiff (35)


Hence, the general solution can be written:

X= C, Coshi^iZ) + Q fihU/^jZ) -^C^ C&$(fi2z)+Cj, Smf^az) (36)

where Cj to C^ are constants to be determined from the boundary conditions.


Combining Eqs. (22) to (25), setting ^~]j -jp ~ 0 gives:
7.2.34

p* EI dz
Substitution from Eq. (36) allows determining "^x^W ^^-^^ ^ <• Next, the
four constants can be evaluated by setting:

At z-0- x-x„ e=®H K=KH ^-K1

At the other end of the shaft section, set:

At 2 ^ 4 ' x-x„^, ©^@„, H,r=H^,H*f K--VC,^„

Thus, the desired relationships become:

(41)

2 I s

(42)

(43)
where:

a,„= (p^€0$h\^pla$X^)/(fi-^^l) (44)

^tH = (fl(^^h\-^fic0iK)/i^hp (45)

^3, = (/S,SmU,-t/S,SihAj/(^,V^,')t (46)

a^,- 2(cosa,-c.sAj/(/S^/St)i' (47)

%», = 6(/S,s;hU.-/S,sm)^)/(/3^/S,')/3'4 (48)

%H = ^^2 ^'^*»^« •+l^f^ihK)ApUfiVfi^^H (49)

a,„ - 6(/^!s;na,-"^a smAt)/(/3>^,')^^i' (50)

(51)

L- 2 - ^ (52)

^^^--m
. ^
''''
X,-^j4 K^f>tL (54)
z •A l (55)

It should be noted that AJU / "tn and n^*, are the static Influence coefficients
neglecting shear deformation. Furthermore, for sufficiently low speeds and suff-
iciently many mass stations, the shear effect can be Ignored and pS'O, Then,
the coefficients ^j„ to ^^„ become equal to 1 and the shaft mass can be lumped
with good approximation at the ends of the shaft section, setting Mt^-O in the
above equations. In most practical cases, this will be accurate enough.
UNBALANCE
BEARING n n^ I BEARING
• L . ^
2 ""3 ^"4
•^ — z»zijs-^q
" ^

Oyfline of Rotor with Locotiofi of Rotor Stations

IfttAj^.an.EntGj,,^

^0, .^^ntl
I.
I
STATION 0 STATION ( n ^ l )

Sign Conftntion for Amplitydei Slope, Binding Moment end Shear Force

Fig. 1. Location of Rotor Stations and Sign Convention for Analysis


7,2,37

0.5 . '0.5

A
A

END 1^OSIT IONS

A
/
n
1•

K
o

'"'"**'**%_
iL
i
i
A
1

u - ^

A
/
^x. ***»«»»». _—...»«« """"^
^ . j t f ^ ^

/A o
f A
/ O
m

K • 1

V
I I •

/
/
/
/
r1
i r » I F™ ,f\ » . ™ , < - ~ , - i - , , p-\ff~rf^s*fi •yrf^

mtUKtflLAL Kt5ULI5
EXPERIi ENTAL DATA
o ® COUPLING END

AA FREE END

0 i\ £\ YiI 16 2D 24 28
SPEED, RPM X 10^

Fig. 2. Unbalance Response of the One-Disk Rotor, End Positions


(There are two sets of data for each measurement plane by using
two different angular positions of the unbalance weights on the rotor)

MTI-2171
0.5 0.5
1 A

A
r-.„.
1 1
"T~~

BEARING POSITIONS
<

$
1 \
©
^
>
4

( L„
/ A W^^ -^^^
" / • ^t^
'x.
/X
^ X•

n
/

r r —

« J f © •
THEORETICAL RESULTS
EXPERI MENTAL DATA
tUUP Liib tNU titm m
A A FREE END 1BEAR
m

0 4 8 12 16 20 24 28
SPEED, RPi X 10^
Fig. 3. Unbalance Response of the One-Disk Rotor, Bearing Positions
7.2.39

m
»

/ o N o
»

J
/
f o
m
ps ^ "
m

.^^^^
I »

o c; •
m
/
^ i.O i ®
/
1

/< )
/
O
I
7
CO

CO
o i
J
m

sn
go. /
/
/

®
CENTERLINE
n i
fi. 1i
.05
*- • '^
J
CO i — ^l ^ ^
oa.
CO
—JH A ; ' 1

Ti ir-rsDi-Tir' h1 n r - o i ii •r n ^ \

o ® EXPERIMENTAL DATA ' 1


1

0.01
8 12 16 20 24 28
SPEED, RPM X 10^

Fig. 4. Unbalance Response of the One-Disk Rotor, Center Position


7.2.40

/ m
5 "*7" ^ - ^ n^

/ o — — •

o / °
c
? LO / >
r.

J
O J
I

CO
1 I€AVY EN[)

0.!5 i 0.5
/ 1
-~-w~~] k
Q /

i rlEORETICAL RESULTS
m / o . E^ PERIiEiTAL DATA
0.
. 11
o
Q.
€0

0.01
0 I\ i1 12 H5 20 24 28
SPEED, RPi X 10^

Fig. 5. Unbalance Response of the Two-Disk Rotor, Weights In-Line, Heavy End Positions
7.2.41

®
m
>
/ " o o o
1.0 ® »
o D
/ ®
m
1
M " /
O
• /

€/5 /
® i
r

CO
Q
OS

go. 1 HEAVY END BEARING


1
1
0.5 i 0.51
O
o_ 1 A
CO
h A _l

-PI i r - / \ r 4 r - T i / ^ ft 1 r s r - m i i •!r r >


incurcLiiuHL nCOULIO j
o ® EXPERIMENTAL DATA

0.01
8 12 16 20 24 28

SPEED, RPi X 10^

Fig. 6. Unbalance Response of the Two-Disk Rotor, Weights In-Line, Heavy End Bearing
Position

MII-2175
7.2.42

.0
X- \ \ <>
__JO__ i) \

m/
X• • \
o o \ ^^^-^ ^ -
/ / ^ •
/° \ / ^ 9
11
- — _
O I.
r
_J A 1•

m • /
KJ
o
I

CO
• / f
CO 0. rFMTFRI IMF
Q
1o 0.5 1 0.5 1
a:
m
tt: ft n
o
ijj
CO XUrAorTII^AI D C C I I I TO
o i ncuni-1 iwHL ncouLij
Q_ o . EXPERIiENTAL DATA
CO

0.01
0 8 16 20 24 28
SPEED, RPi X 10^

F i g . 7. Unbalance Response of t h e Two-Disk R o t o r , Weights I n - L i n e , Center P o s i t i o n


7,2.43

1 1 ®
® (»
o
m 3
o o

J< ' \ L
f
O
^
o ^
»».»_

V
/ \

A
1
>
o ®l

•/
/
/
r o
1
o

1 J
LIGHT END BEARING

/ 1 0.5 i 0.51
/
/ I — "
_
k
_
1 1 '• L_J
/

/ THEORETICAL RESULTS

I o ® EXPERIMENTAL DATA

0 I\ £\ 12 16 20 24 28
SPEED,RPi X 10'

Fig. 8. Unbalance Response of the Two-Disk Rotor, Weights In-Line, Light End Bearing
Position

MTI-2177
l.l.kk

<•

A
. \
#

l \ i _ —
o
o
/o V

\
v^
7
o
c
D

/ 1 iriHT Fwn 1
1
" / 0.5 I 0.5 i
/

. /
ft A
o
/

• / THEORETICAL RESULTS
r o • EXPERIiENTAL DATA

8 12 16 20 24 28
SPEED, RPi X lO'

Fig. 9. Unbalance Response of the Two-Disk Rotor, Weights-In-Line, Light End Position

MII-2178
7.2.45

jf%. O.b
m '>

-^r^^l

/
r
r< »
m
\
\
V
0.5
®/
>
1itAV r tm
4

m
s;7

i /
/
/
^

1
1
if
o J

®/

THEORETICAL RESULTS
o» EXPERIiENTAL DATA

1 1 1 1 ^ J
0 8 12 16 20 24 28
SPEED, R P i X 10'

10. Unbalance Response of the Two-Disk Rotor, Weights Out-of-Line, Heavy End
Position

MTI-2179
l.l.hb

©
t

~v o

©

<>
V •
!y f
^ * * 1 t _
^^ •^-Xs,^^ 3

7 ^ - N
s^
m ^ ^ x^
/
© -^
• ''^'*"^^ w. b
/f « I
- ^
/
• —-H
<
(/
o
05
•/

OJ
• / 1
/ 11L=H¥ 1 l»ilL/ UL.mi^ii'^v 1

/
/
/
/

'

— THEORETICAL RESULTS
o • EXPERIiENTAL DATA

1 1^1 1 1 J
8 12 16 20 24 28
SPEED, RPi X 10^

Fig. 11. Unbalance Response of the Two-Disk Rotor, Weights Out-of-Line, Heavy End
Bearing Position

MTI-2180
7.2.47

„1
@

f'\
1 I ®

O
/ \"• ^ «1
1.0 • 1
m
/ V
f V.^
"J 1
r*®»»-.«
r— ^
/
/
o
I /
0

CO J/ CENTFRLINE.

o
©
7 05

0.5
® /
0.1
CO

i
CO
ixJ
oc

THEORETICAL RESULTS
o• EXPERIiENTAL DATA

1 I 1 1 1
0.0I|
3 ^i i1 i;2 16 20 24 28
SPEED, RPi X 10'

Fig. 12. Unbalance Response of the Two-Disk Rotor, Weights Out-of-Line, Center Positi on

Mri-2181
7.2.48

, ri »

•© • • 1


\
yj /
o 1.0 /o
f
\ .
^ r •
s^\ j t »
J >
ffi
z>
7 __o P—»^
---^^
/
O t
/
1
f
ft 1
CO
•/ 0.5

CO J^X^^^^Sl
?BIT RADIU

O i T
# /
' 0.5
p

/ LIGHT END BEARING


o 1
LLJ 1
CO
1
o
Q_
CO
LLJ
1 r—-i

— THEORETICAL RESULTS
o» EXPERIiENTAL DATA

1 I 1 ...,..±™^ L^ L
8 12 16 20 24 28
SPEED, RPi X 10^

Fig, 13. Unbalance Response of the Two-Disk Rotor, Weights Out-of-Line, Light End
Bearing Position

MTI-2182
1.2.h9

m
>

PV
k
o
» m

e o
<^1)
1
<>

o
\
®
X ^

1 ^ S^ 1 >**^
J
/
/
<
/
f
J

1
%

0.5
n
1, n
m

0.5
/
/ L GHT END
/
/
\ m

'

THEORETICAL RESULTS
o ® EXPERIiENTAL DATA

1 1 1 1 1
8 12 16 -20 24 28
SPEED, RPi 110^
Fig. 14. Unbalance Response of the Two-Disk Rotor, Weights Out-of-Line, Light End
Position

MTI-2183
7.2.50
A

1
END POSITIONS
0
I i 0.33
0 t 0.33
h I ' •'

iO.33
\

\
A
(
i\
-
T \

ml
A
V
o
<^

THEORETICAL RESULTS
/
• EXPERIiENTAL DATA
2 o » COUPLING END
/
A A FREE END
/
i
<

J ^\

J
A

0 i\ iI 12 16 20 24 28
SPEED, R P i X 10'

Fig. 15. Unbalance Response of the Three-Disk Rotor, End Positions


7.2.51

1 1 1 1
A
BEARING POSITIONS
1 1 1 1

i, 0.33
f 0.33
" A - A

.
1( 133
1
&
f THLORtTICAL RESULTS
EXPERIMENTAL DATA
o® COUPLING END BEARING

1, A A FREE END BEARING

1
4

©

/ \
J X•\
M
^

r1 n V
V
f b.
\ \

1 /
I
^
^

'/I
/
/ -'"""—

0 4 £\ 12 HD 2D 24 28
SPEED, RPi t 10^

Fig. 16. Unbalance Response of the Three-Disk Rotor, Bearing Positions


•I
1 1 1
CENTERLINE
1 1

®
1 0.33 t 0.33
K A
o
I 0.33
j
-™ THEORETICAL RESULTS
L_ ^ o ® EXPERIMENTAL DATA
,

(f

ml \

\ I[s
i
/i
>
\ .
1 X.
J
I — _

1
1'

0 A
J i^ 12
SPEED, R P i K 10'
16 20 24 28

Fig. 17. Unbalance Response of the Three-Disk Rotor, Center Position


00
l.l.l-i

>

BEARING AND SEAL REQUIREMENTS FOR LARGE


LIQUID-METAL COOLED REACTOR SYSTEMS*

by

Robert W. Dickenson

Atomics International
Canoga Park, California

SUMMARY

Bearing and seal technology is well advanced in the components to which


it has been applied I So far, it has been concentrated primarily in the
mechanical sodium pumps; an order of magnitude extrapolation in pumping
capacity will be required for prospective plants, which in turn requires
technology extrapolation and extensive tests to assure satisfactory per-
formance. Some additional applications of bearings and seals are possible
if they are economically justified, although at least partially satisfactory
substitutes are currently available.

INTRODUCTION
Fast breeder reactors of power levels of 1000 Mwe or greater are presently
considered to be necessary for the continued conservation of nuclear
resources in an expanding nuclear power economy. Component requirements
in the large future systems transcend any experience now available; the
largest sodium cooled reactor built to date has a design power of
300 Mwt, with components sized to meet this power generation capability.
Therefore, an order of magnitude extrapolation in component size for
power generation at about 2500 Mwt is a near term objective.

Sodium, because of its high chemical reactivity, and its rather poor
lubricating qualities is not suitable for anti-friction bearings. Conse-
quently, sodium systems have demanded bearings lubricated by the

*This work was supported by the U.S. Atomic Energy Commission under
Contract AT(ll-l)-GEN-8
8.1.2

working fluid and rotating seals to retain inert gas over the sodium wher- y
ever mechanical shafting had to be introduced into the heat transfer
systems, almost since the inception of sodium cooled reactors. Sodium
also becomes highly radioactive in the primary circuit. While this does
not affect the bearing performance, it does introduce stringent require-
ments for reliability, accessibility and repairability. In small plants,
the requirements could sometimes be dodged by devices such as frozen
sodium shaft seals which permitted conventional bearings to be used out-
side the sodium, with varying degrees of success. However, with the
advent of even the moderately large plants constructed to date, it has
become apparent that at least for pumps the impeller and shaft must be
stabilized by sodium lubricated bearings. This automatically requires
the use of a rotating shaft seal to retain inert gas over the sodium in the
pump casing; we are faced again with a large extrapolation from current
practice,

CURRENT APPLICATIONS

Bearings

The largest sodium pumps built to date have been installed in the
Enrico Fermi Fast Breeder Reactor, with a capacity of approximately
11,000 gpm. These have been operated quite satisfactorily, indicating
that the principle of sodium lubricated bearings is a sound one. The
bearing dimensions for these pumps are 12-in. diameter and 14-in.
length. The six pumps installed at the Hallam Nuclear Power Facility,
of 7500 gpm capacity and identical bearing size, have operated equally
satisfactorily for a total of 70,000 hours on all pumps.

The bearings in sodium main circulating pumps to date have been


hydrostatic, supporting only the radial loads imposed on the shaft. High
pressure sodium supplied to the "pads" is furnished from the discharge
of the pump, with 1 to 4% of the total pump flow being bypassed through
the bearing to the suction side of the pump. Bearing sodium is relieved
into the free surface volume of the sodium above the pump impeller and
flows back to the suction side of the pump through an overflow line and
pressure balancing "weep holes" in the impeller. Radial loads of up to
120 lb force have been sustained by these bearings, which operate in
the turbulent regime. No attempt has been made as yet to sustain thrust
loads by such bearings; these are typically absorbed by roller or duplex
ball bearings located above the rotating inert gas seal, where they are
available for inspection and receive ordinary oil lubrication. Journal
surfaces are hard faced to prevent galling during startup and shutdown
with evident success; in the case of the Hallam pumps, inspection after
250 starts and stops showed no measurable wear and only slight polishing
of the surfaces. Experimental evidence has been adduced which indicates
that bearing stabilization and "lift off" from the rubbing surfaces takes
place after less than one revolution of the shaft. Because of the success
of this design on the pumps constructed to date, this type of bearing will
undoubtedly be at least one of the reference designs for the 60,000 to
120,000 gpm pumps now under consideration. There are, however,
obvious unknowns in extrapolating this bearing performance to the rather
larger shaft sizes envisioned. The design parameters of these larger
pumps have not yet been firmly fixed. It is reasonable to expect, how-
ever, that bearing journal diameters of at least 20-in. will be required.
Shaft rotating speeds will probably not be substantially different from the
500-1500 rpm commonly utilized now, but it is obvious that the lineal
velocity at the bearing surface will be substantially higher due to the
increased bearing diameter. These pumps will be used in commercial
power plants, where any loss of efficiency means an increase in parasitic
plant load, which in turn means internal consumption of power which
cannot be sold at a profit. Efficiency is thus a matter of great interest,
particularly when one is considering pump drivers of 5-10 Mw in power.
Bearing design thus assumes one of the critical roles in plant design and
pump selection, which must be done on the basis of economics and
reliability. A high by^jass flow and/or a high bearing drag will contribute
a significant efficiency loss to the plant with a measurable economic
penalty. There are other types of pumps which can be utilized for moving
sodium which are currently less efficient than mechanical pumps, but
whose other potential advantages may prevail should the mechanical pump
efficiency fall undesirably low. Therefore, it should not.be assumed that
free surface mechanical pumps with sodium lubricated bearings are the
only possible choice regardless of their mechanical design and efficiency.
For instance, small canned motor pumps have been used in sodium to a
limited degree; these utilize hydrodynamic bearings and have apparently
been satisfactory although their cost has been quite high compared to
shaft sealed overhung impeller pumps which can perform the same service.
Bearings are probably not the key to reducing cost in this case; electrical
insulation is a more likely candidate for cost reduction.

Seals

An inert gas blanket must be maintained over the sodium pool within
the free surface pumps. This gas is normally argon, in fast reactors,
which is maintained at 1 to 5 psi over atmospheric pressure, to permit
outleakage of argon rather than inleakage of air in case of seal failure.
In addition to minimum inert gas leakage, the rotating seal must not per-
mit any contaminants to enter the sodium stream. Oil lubricated face
type seals must be carefully designed so that the lubricant is trapped
before it can leak down the pump shaft, and the oil volume should be
arranged such that the traps can contain any conceivable amount of
leakage. Carbon is quite deleterious to sodium systems, carburizing
and embrittling thin stainless steel sections such as fuel cladding, and
its introduction into a sodium system must therefore be avoided. It is
possible to find non-carbon bearing lubricants; the effects of their
decomposition products in sodium has not, to my knowledge, been com-
pletely explored. The simplest course is therefore to assure that
potential contaminants cannot escape from the seal into the sodium.

Rotating seals utilized in large mechanical sodium pumps to date


have been carbon faced, oil lubricated, operating on shafts up to 6-in,
in diameter. Again, a substantial extension of technology is required
when considering this type of seal for shafts up to 10-in, in diameter.
Present seal technology is demonstrated at lineal speeds up to about 30,000
inches per minute; whether this experience, technology and manufacturing
practice for flatness, finish, etc., can be extended by factors of 2 to 3
satisfactorily has not been established. Complete reliability for at least
10,000 hours of operation at speeds of 1500 rpm is required, as the seal
is essential to pump operation, and the reactor system cannot operate
without pumps. It will be necessary to perform both analysis and
extensive tests to ensure that this critical component meets the require-
ments of economy, reliability, and tightness to gas and lubricant leakage
which are required for performance in the central station fast breeder
reactors of the future.

POTENTIAL APPLICATIONS OF BEARINGS AND SEALS

Bearings

Some possible additional applications of sodium lubricated bearings


come to mind, providing the technology permits. It should be understood
that these are not necessarily recommended applications, and that reason-
ably suitable equipment now exists. The examples cited must offer sig-
nificantly improved reliability and/or cost advantages to be seriously
considered. The first of these is in turbine type flowmeters, completely
immersed in the sodium stream. These are not now practical primarily
because of the bearings, which cannot be lubricated by anything but
sodium. There is no source of high pressure sodium for these small
bearings as in the pump case; a hydrodynamic bearing might possibly
be suitable if the cost and complexity can be reduced to where such a
flowmeter would be competitive with the electromagnetic type (which
suffers from lack of accuracy in large pipes) and the differential pressure
type (which suffers from lack of accuracy due to imperfect pressure
measurement), A considerable amount of ingenuity will also be required
in the electrical in'sulation and connections to a turbine flowmeter in
sodium, even if the bearing problem is solved.

It is also possible to conceive control and safety mechanisms which


could operate at reasonably high speeds in sodium, providing adequate
bearings were available. As it is, all bearing surfaces are normally
kept out of sodium and sodium vapor atmosphere, with the possible
exception of surfaces which slide or roll quite slowly over one another.
Sliding and rubbing under sodium, or in sodium vapor, has proven to be
most unrewarding to date, although development along this field is
8,1.6

continuing. A "canned motor" drive with sodium lubricated bearings is


conceivable, although many other problems such as electrical insulation
¥
would face the designer.

Seals

Any time a sodium system is penetrated, it is necessary to utilize


a seal. Normally, the seal must retain inert gases containing sodium
vapor. The main sodium coolant pumps are the prime examples. Control
rod drive mechanisms, which usually penetrate the top of the shield,
are excellent candidates for improved seals. Currently, the problem is
sometimes dodged by isolating the operating mechanisms from the
reactor atmosphere by a bellows which operates only in (vertical) linear
motion. This is not always desirable, due to the questionable integrity
and lifetime of thin-walled bellows with large extension ratios. This
problem has also been dodged by containing the whole control mechanism,
including the absorber column, in a "thimble" thus isolating the com-
plete unit from sodium and cover gas. This is wasteful of neutrons, and
is often rather clumsy in a small, high power density, reactor core.
Therefore, a reliable rotating seal, operating at modest speeds up to
perhaps 100 rpm, which can be depended upon to permit the escape of
cover gas containing radioactive sodium vapor and maintaining close
alignment throughout its lifetime, may well find application in drive and
safety mechanisms for sodium cooled reactors.

So far, sealing against sodium directly has been most successfully


accomplished by a frozen sodium annulus, such as in valve stems.
Cooling is either by air or a liquid jacket. This method of sealing does
introduce a necessity for periodically exercising the valve stems to
break up sodium oxide deposits which may form. Also, of course, some
small amount of sodium oxide is deposited on the shaft each time the
valve is operated, thus making operating force erratic. So far, this has
not proven difficult to handle, requiring only a sufficiently energetic
valve operator. These are normally electro-mechanical or pneumatic.
An economical seal, with constant force characteristics and which can be
depended upon to reliably seal high temperature sodium from the atmos-
phere in translation and/or rotation would be highly desirable.
J
8.2.1

DESCRIPTION OF A LARGE SODIUM PUM> CONCEPT


FOR FUTURE SODIUM COOLED POWER REACTORS

by

Donald R. Nixon
Westinghouse Electric Corporation
Atomic Equipment Division
Cheswick, Pa.

SUMMARY
A review of sodium pump and sodium cooled reactor technology was made by
Westinghouse in 1963 under U.S. AEC Contract No. AT(30-1)-3123*. A purpose
of this review was to establish criteria and limits for the sodium pumps which
will be required for large (1000-1500 MWe) sodium cooled power reactors. This
study indicated possible pumping requirements of 5000-10,000 hydraulic horse-
power per pump.

DISCUSSION

The conceptual design shown Figure 1 depicts a 6000 HP electric motor coupled
to a vertical pump shaft. The shaft extends down through a five foot thick
shielding plug into the pump tank, A free surface of sodium is maintained in
the pump tank by gravity or cover gas pressure balancing. By providing an inert
gas cover above the sodium level in the tank, ingress of air is prevented and
the requirements of the pump shaft seal (located just above the floor level)
are reduced to that of sealing 5-50 psig gas pressure and sodium vapor. Radial
baffling within the pump tank is provided to prevent vortexing of sodium around
the shaft. A labyrinth or series of conical baffles is shown above the free
surface to limit the passage of sodium vapor and to reduce convective heat transfer
in the cover gas area. A sodium lubricated radial bearing is shown just above the
pump impeller at the lower end of the shaft. The upper radial and thrust bearings
are located above floor level and above the gas shaft seal. They are shown as
conventional oil lubricated bearings.

The hydraulic or pump end shown is a double suction pump in which the sodium is
drawn into the top and bottom of the impeller and discharged through a diffuser
into the pump casing or volute. The pump internals (impeller, diffuser, lower

* see bottom of Page 8.2.3 for report titles and numbers.


bearing assembly, and baffling) can be withdrawn vertically from the pump tank
without requiring personnel to work below the shield floor level.

The leading particulars for the pump hydraulic design shown are:

Fluid - 800-1200°F Sodium


Flow - 80,600 GPM
Total Head - 300 Ft.
RPM - 890
Specific Speed - 3500
Hydraulic Efficiency - 887o
NPSH Req'd. - 62 Ft.
Impeller O.D. - 42 Inches
Suction NOZZIBJI.D. - 40 Inches
Discharge Nozzle,I,D. - 36 Inches

The overall length of this unit is approximately 45 feet and the diamter of the
pump is five feet.

This design is strictly conceptual and major development areas include:

1. Shaft Seals - Development of large diameter seals capable of


sealing 5-50 psi cover gas and sodium vapor with little or no
maintenance.

2. Bearings - Development of large diameter sodium lubricated radial


bearings for 800-1200 F sodium service.

3. Pumps - Development of high capacity pumps with low NPSH requirements.

Other areas which will require detailed analysts are:

1. Thermal Studies - To provide information necessary to design for


thermal shock and expansion.

2. Dynamic Vibrations Analysis - To provide information to assure


satisfactory shaft and rotor design.
8

3. Fluid Stability Studies - To provide information to assure sodium level


control and prevention of gas entrapment in the sodium stream.

4. Mechanical Design Considerations - Obtain information for proper


layout with respect to the reactor and other nuclear considerations,
and to insure case of installation, assembly, and disassembly.

5. Materials Selection - To assure structural integrity and long life


for motor and pump components,

6. Electrical - To assure proper motor sizing and speed regulation with


associated nuclear considerations for motor lubricating and cooling
systems.

7. Instruments and Control - Selection and location of the various


instruments required for monitoring temperatures, pressures, vibration,
liquid level, etc. af pump and pump auxiliary systems.

Westinghouse is presently performing parametric and design studies under U.S. AEC
Contract No. AT(30-1)-3762 directed towards development of Engineering data for
large mechanical centrifugal sodium pumps.

Due to the size and speed requirements envisioned for the large sodium pumps of
the future, development of reliable turbulent flow sodium bearings and gas seals
will be necessary.

Reports published in the study under U«S. AEC Contract AT(30-1)-3123:

WCAP 2347 - Sodium Pump Development and Test Facility Design

WCAP 2255 - Survey of Sodium Pump Technology


00

H
OQ
©
m
o
IM^^ri: 3 S' ,'

f " far/ ., \?iv--% m.f*^:/


xs

n
CD
8.3.1

PROCESS FLUID LUBRICATED BEARINGS FOR HIGH-


TEMPERATURE GAS-COOLED REACTOR CIRCULATORS

by
J. Yampolsky
General Atomics Division of General Dynamics Corp.
San Diego, California
and
D. F. Wilcock
Mechanical Technology Inc.
Latham, New York

ABSTRACT
The development of a new gas cooled power reactor, has required the engineering
of a new steam-driven, high-pressure, helium circulator. Because of the
simplifications possible in seals and lubricant systems, water lubricated
bearings were selected.

A shrouded step design was chosen for both journal and thrust bearings. These
designs offer the advantages of a large tolerance to wear, no moving parts,
and the flexibility of running as hybrid or even hydrostatic bearings if desired.

An approximate solution, accounting for turbulence effects (the Reynolds Number


may exceed 25,000) has guided initial design. Some results from a computer
solution for the thrust bearing taking into account inertial effects as well as
turbulence are described.

DISCUSSION

The 330 MWe Power Plant being constructed by General Atomic for the Public Service
Corporation of Colorado has the entire primary circuit consisting of the reactor
core, steam generators and circulators, contained entirely within a prestressed
concrete pressure vessel. The output power of the plant is produced by a con-
ventional 2400 psi 1000 F/IOOO F reheat steam turbogenerator which is located
together with the remainder of the associated power plant equipment in a separate

W turbine building. The only penetrations of the PCRV required are those for
power production; feed water in, high pressure steam out, cold reheat steam in,
and hot reheat steam out. The concept of the series steam turbine driven helium
circulator was conceived by General Atomic to best provide the means for primary
helium flow in the reactor power plant. Each of the circulators consist of a
single stage axial flow steam turbine that is in series or tandem with the power
producing steam turbogenerator. These driver turbines are located in the cold
reheat line between the discharge from the high pressure element of the turbo-
generator and the reheater. Figure 1 shows the series steam turbine driven
circulator installed in the bottom head of the PCRV. The machine is oriented
vertically and its housing forms the closure to the primary gas circuit. The
arrangement of the elements comprising the circulator is shown in Figure 2. The
compressor and turbine rotors are overhung from a center bearing housing on
opposite ends of the shaft. The bearing housing which contains the bearings and
seals is supported by the conical section that forms the closure to the primary
system.

The choice of the bearing type and lubrication system is strongly influenced by
the operational and system requirements. The loads imposed on the radial bearings
are those resulting from unbalance since the orientation of the machine is vertica
while the thrust bearing loads result from the pressures acting at each end of the
machine and the axial reactions of the turbo machine elements. It should be borne
in mind that although nuclear plants are, in general, base loaded, they require
starting and stopping and operation under emergency conditions. The thrust loads
for all of the required conditions are identified in Figure 3, where the speed
(plant load) is the abscissa and the thrust loads and pressure are the ordinates.

Furthermore, these large variations of pressure of the two working mediums,


steam and helium, the mechanical requirements of clearances of at least one
thousandths of an inch per inch of shaft and the variable speed operational
requirement determine the need for hydrostatic lubrication for at least the low
speed part of the operating range and emphasize the problem of insuring effective
sealing of the fluids in the machine. The fluids in the machine must be kept
separate and the only proven and effective method of sealing for long periods
of operation is to allow leakage in a controlled fashion. One, therefore, resorts
to a buffer gas leaking into the process and into a scavenge chamber into which
8.3.3

the lubricant also leaks. The lubricant and buffer gas are then separated in
the accessory system and returned to their respective systems. Moreover,
since the pressures in the two process fluids vary considerably from each other
at part load, large flows of buffer gas would be required or each of the bearings
would have to be pressure referenced to the process it is contiguous to, with
resulting leakage flows in the lubrication system between the two pressure systems.
If, however, one of the process fluids is used in the lubricamt, a single pressure
system referenced to the other process fluid is feasible with a considerable sav-
ings of equipment and complication as compared to the two-pressure lubricant system
and a two buffer gas system. The choice of water as the lubricant allows for this
simplification.

As was pointed out earlier, hydrostatic bearings are required by the operating
conditions so that the boundary lubrication proeprties of oil are not necessary
for startup or shutdown, and it is possible to always have large clearances
separating the possible rubbing surfaces. Another contributing factor to the
choice of water as the lubricant, is that redundancy of the lubrication supply
is easily achieved by the use of the feed water system as an emergency lubricant
supply.

Three types of thrust bearing were considered, a tilting-pad bearing, a pocketed


hydrostatic bearing, and a shrouded step bearing. Because of the difficulties
associated with providing hydrostatic starting for a tilting-pad bearing, as well
as its complexity, the tilting-pad design was not considered further for the thrust
application. Furthermore, since the shrouded step design could provide a large
measure of hydrodynamic lift performance, and in addition could operate as a
hydrostatic bearing at low speeds, it was selected as the design to be investi-
gated further.

This paper describes the preliminary design for the shrouded step thrust bear-
ing. It is discussed here because it is the critical bearing in the design
because of power loss and thrust capacity considerations.
BEARING DESCRIPTION

The shrouded step bearing is shown diagramatically in Figure 4. A number of pads


of the general shape shown are arranged on the annular thrust bearing surface.
The step is depressed an amount (h„-h-|) below the otherwise uniform top surface
of the thrust bearing. Note that a feed is provided at the entrance to the step
and that a drain is provided between each pad and its neighbor. In operation,
fluid pressure is developed in the step to support load.

Approximate Analysis

Several assumptions were made which permitted a closed form solution to be obtaine
In addition to the usual assumptions involved in establishing Reynold's equation,
the bearing was assumed to be a linear strip, thus ignoring the curvature of the
boundaries and the difference in length between inner and outer edges. Further-
more, pressure across the step (in the radial directiion) was assumed to be con-
stant, so that a one dimensional solution could be obtained. Finally, the corner
areas beyond the step (Bxe) were ignored both for their contribution to flow and
their contribution to load capacity.

The basic differential equaticns were the same by summing flows into and out of the
strip of length dx running across the bearing. Both shear flow and pressure flow
were considered in the direction of motion, and end flow out of the ends of the
strip was assumed to be due to pressure flow only.

Because these bearings were to operate in the turbulent regime, with Reynolds'
numbers as high as 50,000, provision was made to introduce an effective viscosity,
corrected for the influence of turbulent on viscosity.

The resulting differential equation was

^ - M — ^ ^ = 0 (1)
dx^ ^^ ^^z ^2
where
|a is the effective viscosity in the x direction over the step and |i
X-ii z
is the effective viscosity in the z direction over the lands. u^
•' xl
is the effective viscosity in the x direction over the lands.
8.3.5

Equation (1) applies along the step. Assuming a linear pressure drop over the
lands, Equation (1) maybe integrated and with the appropriate boundary conditions,
giving the following relationship for the load:

1 e . ,
W = + 2 i s^" h ^
(2)
^ + ^ sinh A ^ + ^ cosh

where
3 -, 1/2
2E! / ^ I \ I \
^ = bL (3)

H = (4)

|^6V(h2-h^) en ,/p.h^ J - cosh TV - HiL sinh 7L


jr = (5)
sinh A + H iV cosh TV

Equation (2) is conservative in that it does not include the load carried by
pressure in the feed groove itself or over the land between the drain groove and
the feed groove.

Calculated Performance

It was desired to have this bearing fully operable over the full speed range from
30 percent to 100 percent of full speed. Bearing proportions were analyzed there-
fore at the worst load condition at 30 percent speed.

The effect of the width of the side shroud is shown in Figure 5, holding the total
radial width of the bearing constant. While the maximum load is obtained at a land
width of about 0.3 inches, the fluid flow through the bearing increases rapidly as
the land width is decreased. The dotted curve in Figure 5 shows the load per unit
flow, perhaps a better measure of overall bearing performance. Therefore, a value
of land width of 0.475 was selected as a compromise between load capacity and flow.

The step depth is also a critical design parameter. If the film thickness, h^, is
held constant the step depth is gradually increased, the load carried by the bear- •I
ing increases rapidly at first then reaches a optimum value at approximately 5 times
the film thickness. The optimum point is a function of feed pressure, film thick-
ness and the other design variables. A value of 5 was selected for this design,
corresponding to the optimum at the 30 percent speed point with the desired minimum
film thickness of .002 inches.

The length of the land beyond the step, e, cannot be optimized with these approximate
equations. These equations indicate an increasing load capacity as small e is
increased. However, due to the approximations made in eliminating the corner flows,
it is very doubtful whether this is the case. For this reason, small e was chosen
at an arbitrary value that would provide reasonable flow rates, and was roughly
comparable with the side land width.

The pressure generated at the step is shown in Figure 6 for two supply pressure.
Here the step depth is kept constant at 0.012 inches and the h /h^ ratio is varied
by varying the film thickness h^ . It is seen that the pressure at the step in-
creases rapidly with h„/h^ for the low supply pressure. This calculation, made
at 110 percent speed, shows one evidence of the hybrid or partically hydrostatically
nature of this bearing. At small values of h„/h^ the step pressure is very little
greater with a high supply pressure than with a low supply pressure, suggesting that
pressure drop along the step is a strong factor when a high supply pressure is used
in the feed groove.

In Figure 7 the pressure at the step is shown as a function of speed, with both the
step depth and the bearing film thickness held constant. With a low supply pressure
the pressure at the step increases regularly with speed at a rate faster than linear.
For a high supply pressure, 250 psi, the pressure at the step is nearly equal to
the feed pressure at 0 speed drops to a minimum of about 1/2 feed pressure at 4000 rpm
and then increases again. This again is evidence of the quasi-restrictor action
of the step towards lubricant flow in the bearing.

The overall performance of a 5 x 9-3/4 6-pad thrust bearing is shown in Figure 8,


as a function of speed. The load curve is that imposed by the machine performance,
and it should be noted that this may differ from the final load curve reported in 4
8.3.7

the first part of the paper, because in the course of design the load curve has
shifted. The solid curve for film thickness at an inlet pressure of 250 psi
shows that at no time does the bearing go below the desired minimum film thick-
ness of .002 inches. The solid curve marked flow in gallons per minute gives
the flow through a single pad. Note that the flow is very high at full speed and
again is very high at very low speeds.

Because it was noted that high flows occurred where the film thickness was much
greater than the minimum required, the bearing was redesigned to include a flow
controlled restrictor in series with the bearing. Using a higher, but available,
supply pressure of 600 psi plus a restrictor, the film thickness and flow rate
curves shown by the dash lines were obtained. By this means, the flow rate was
kept nearly constant over the entire speed range, and the film thickness was
maintained above the desired minimum value from 30 percent speed to full speed.
A single restrictor in series with the six bearing pads was used rather than an
individual restrictor per pad. Thus the restrictor served primarily as a constant
flow control device, reducing the speed pressure at the bearing in the speed
ranges where large flows would otherwise be obtained.

More Exact Analysis

While the approximate solution discussed above takes account of turbulence effects
in the bearing, it ignores the influence of other inertia affects. These inertia
affects can be expected to manifest themselves in three ways. These are Bernoulli
pressure drops at steps or entrance areas, ram effects at entrance, and centrifugal
effects. The scale of these effects can be large in bearings of this size. For
example, the Bernoulli pressure drop from an entrance groove into an area with an
average velocity equal to half the runner velocity, is more than 100 psi. For
these reasons 5 a more exact analysis is underway in which these inertia effects
are accounted for in every point in the film. The solution is realized on a
digital computer in two dimensions (R,p) using an iterative technique.

The inlet, exit and transition boundary conditions have been carefully considered.
At a transition such as the step, conservation of total pressure has been main-
tained. At the exit boundaries, the static pressure in the film is made equal
to the external ambient pressure, on the assumption that there is no sensible
recovery of the velocity head possible. At the inlet, no velocity head correction
is made on the assumption that the fluid exiting from the land between the drain
and the feed groove into the feed groove forms a boundary layer which increases
in thickness due to energy imparted by the runner as it traverses the feed groove.
In fact, one can hypothesize everything from a Bernoulli pressure to a full ram
pressure rise at the entrance to the step. Until experimental information is
available, the compromise assumption of no pressure change is made.

A few results from this program are now available. Figure 9 shows a pressure pro-
file along the circumferential mid plane of the bearing for a Reynold's number of
2500 and for a step depth equal to the film thickness. In this representation,
the feed groove is at Station 3 and the step at Station 8. The drain groove is
at Station 0, 10. The lower, solid, curve is the solution with no inertia terms
but including the turbulent effects. The upper curves show the solution with the
inertia correction, including the sharp drop at the step. It can also be that the
values obtained on the first iteration are quite close to the final converged
solution. The centrifugal affect is shown in Figure 10. Here under the same
conditions as in Figure 9, one can compare the inner half profile with the outer
profile. Perhaps surprisingly, the outer half profile is lower than the inner
half profile, indicating that the influence of the longer leakage path overrides
the centrifugal effect. The influence of Reynold's numbers is shown in Figure 8,
in which the dimensionless load is plotted as a function of Reynold's number.
Note that while the dimensionless load appears to decrease as Reynold's number
increases, the true bearing load will increase with Reynold's number since it is
proportional to the square of the angular velocity. Figure 11 shows the curve
with no inertia as the lower solid line^ and the first term correction as the
upper dash line. The middle curve showing the conversion solution has been ob-
tained up to a Reynold's number of 2500, and appears to be reasonably extrapolatable
beyond that point. Thus the first term correction may serve as a guide at Reynold's
numbers for which full conversion cannot be obtained.
8.3.9

PLANNED PROGRAM

An experimental program for the evaluation of this bearing is planned to begin


shortly. Figure 12 is a photograph showing the thrust and journal bearing. The
displacement probes for the measurement of axial and radial clearance are shown
as are the static pressure tappings in and around the bearing pocket. The same
type of probes are used to measure speed and are seen mounted in brackets which
protrude in front of the thrust bearing and over the outer diameter of the thrust
runner which is not shown. The inlet ports in the pockets of both the thrust and
journal bearings are observable in the photograph.

It is planned to carry out the experimental program on a test rig in which the
inertia of the circulator rotor is simulated. In fact, the actual shaft and discs
will be used. This experimental work will provide information on the bearings,
seals and rotor dynamics. Figure 13 shows the contemplated measurements. There
will be five planes in which the radial displacement of the shaft relative to the
stator and the axial displacement of the thrust runner relative to the thrust
bearing will be measured. In addition, the static pressure profiles in one of the
pockets of both the thrust and journal bearing will be measured. In addition,
the flows, temperatures and pressures in the seal and lubrication system are to
be observed.

CONCLUSIONS
The step bearing appears to have particular advantages for this installation which
requires a bearing capable of long life with no attention for 10 years or more
while operating in a high pressure water environment. The step bearing offers
the decided advantage of excellent hydrodynamic performance in addition to capa-
bility as a hydrostatic bearing. Because of the relatively large step depth,
and the small rate of variation in load capacity with step depth when operating
near the optimum point on the step depth in the load versus step depth curve,
the step bearing can be expected to be very insensitive to wear. Thus, in the
bearing under discussion here, wear of one to two mils can be tolerated without
seriously affecting the bearing performance. Although inertia effects can play
a major role in the load capacity of this bearing when operating at high Reynold's
numbers and the nature of the boundary condition at the inlet of the step is an
unknown area which can benefit from experimental investigation, it is possible to
meet most any exigency that might occur through a modification of the hydrostatic
component of this type of hybrid bearing.
SERIES STEAM TURBINE CIRCULATOR
PCRV INSTALLATION

.DIFFUSER

COMPRESSOR
STATOR•

.COMPRESSOR
ROTOR

PCRV LINERv

STEAM
TURBINE
CASING-

-PRIMARY
CLOSURE

WATER
INLETS

Figure 1: Circulator Thrust Load, Helium Pressure


and Steam Pressure as Functions of Speed
SERIES STEAM TURBINE CIRCULATOR

COMPRESSOR COMPRESSOR INLET 'PCRV LINER


DIFFUSER

PRIMARY CLOSURE

STEAM TURBINE CASING

COMPRESSOR STEAM "PELTON


STATOR TURBINE^ WHEEL
ROTOR
'COMPRESSOR
ROTOR

Figure 2: Series Steam Turbine C i r c u l a t o r ,


PCRV I n s t a l l a t i o n
CIRCULATOR THRUST LOAD,HELIUM PRESSURE,AND
STEAM PRESSURE AS A FUNCTION OF SPEED

INLET STEAM /
8000 — 800
PRESS.
\
INLET He \
y^
6000
PRESS. \
> -
^\ .——
7—z
y^ — 600
^ ^
O /

7^
START-UP
THRUST ^OUTLET
4000 STEAM 400
CIRCULATOR PRESS.
DOWNWARD
PRESSURE
AXIAL \ EMERGENCY (PSIA)
THRUST DEPRESSURIZEXL 2 0 0
(LBS) 2000
DOWN y \ THRUST \
THRU ST
\ \
0 \ ^>
\
11 AXIAL THRUST
DURING NORMAL
^A
UP \
THRUST OPERATION
2000 200
\ -

20 40 60 80 100
% SPEED

Figure 3: S e r i e s Steam Turbine C i r c u l a t o r


8.3.13

Drain
Tied

Figure 4: Schematic of Shrouded Step Bearing

P
400

300

W
(LBS)
200

5 X 93/4x6 PAD THRUST


L + 2b-2.375
100 A-0.012 h =^0.002 —
N-2865 RPM

0
0 05 1.0
b

Figure 5: Load Versus Land Width


600
PRESSURE AT STEP
^
5 X 9V4 X 6 PAD THRUST
500-
" A = 0.012 WATER
N= 10,600 y ^

400

PS!
300 Pi == 250/
A •20

200

/
A
/
/

100
/
^

Q
0 2 3 4 5 6
h2/h|

Figure 6: Pressure at Step Versus Step Depth Ratio


400
PRESSURE AT STEP
5 X 93/4 X 6 PAD THRUST
A-0.012 WATER h^-0.005 /

300

Pi = 250
^ /
PSI
200
"^ Pi = 20

100

i 1 f 1 l i l t 1 J I ! .
0,
0 5000 10,000 15,000
RPi
Figure 7: Pressure at Step Versus Speed
8.3.1

40 10
5x93/4 x6PAD THRUST
h,p;=250
A = 0.012 WATER
w
KIPS

30

h
MILS
Q
GPi

20 0

10

_ - . - » » — ^ ' ^ ^ ^ p . = 6 0 0 + RESTRICTOR

0
0 5000 10000 15000
RPM

Figure 8: Film Thickness and Flow Versus Speed


0.30 00

U3
Re-2500, R/C-IOOO, Pg-0.0844
A / C - l , R-0.5125
0.25 (APPROXIiATELY AT THE MIDDLE
OF THE POCKET)

0.20
3

ICL

4 6
NUMBER OF STATION

Figure 9: Computed, Normalized Pressure Profile in


Direction of Motion
0.20

/ NNER HALF ANNULUS


/ 1 1
0.16
^ 'OUTER HALF ANNULUS
f i l l
^

0.12 \
\
3

^
ICL • • - « - . ^
0.08
NO'INERTIA^v X .
__. (BOTH HALVES) O c 1

^ ^ ,
0.04
^TfP
0 1 cr
Ota
•• LAN!)J \ \
,

\
0
0 I 2 3
NUMBER OF STATION

Figure 10: Computed, Normalized Pressure Profile in


Radial Direction
— 1 OS

V X ^X
^
V
\
\ ^ N_
\ \ _. laiiTi 1 1 s
S»w.
\ »«.. V^ %!, wn n 1ST 1 trtivi turtrttt 1 luiM
/
NO INERTIA^ "X, X.*^' ^**^^^»^^^^ /
^^
^"•^Z^ " " " ' ! . ' - - ^»» « "" «»

CONVERGED UP TO Re= 2500

10' 10^ 10^


Re =

Figure 11: Normalized Load Versus Reynolds Number

i|^2275
00
Figure 12: Journal and Thrust Bearing w
N3
MTl-2264
SNSTRUMENTATiON BEARING TEST RIG

, V}P' TAPE RECORDER


P-2'fi){f,tA) CH 13

EXTERNAL
RESTRICTOR

^ - T A P E RECORMR
I CH. 8 4.9

- -OSCILLOSCOPE

THRUST BEARING
PRESSURE. TAPPING

-STEAM TUHBINC
INERTIA DISC

— TAPE RECOWJER
CH. 10 & 11

i/JP\ 1___JT^
V i ^ 2 4 PT. V l y 12 PT.
RACK RACK
MTO MTO

NOTES
(, TWO DISPLACEMENT PROBES ARE INSTAU-EO

90" APART \H SAWE PLANE.


Z P M 5 I N S T A L L E D OM T H R U S T COLLAR
leo'fftOM pi-ts.

UPPER JOURNAL BEARING


PRESSURE TAPPING

Figure 13: Instrumentation for Bearing Test Rig


^

REPORT DISTRIBUTION


DISTRIBUTION LIST

Contract No. AT(30-1)-3363

U.S. Atomic Energy Commission U.S. Atomic Energy Commission


New York Operations Office Division of Reactor Development
376 Hudson Street and Technology
New York, New York 10014 (2) Washington, D.C. 20545
Atten: M.A. Rosen, Asst. Dir.
U.S. Atomic Energy Commission for Reactor Engineering (1)
Division of Reactor Development
and Technology U.S. Atomic Energy Conmission
Special Technology Branch Division of Reactor Development
Washington, D.C. 20545 (3) and Technology
Washington, D.C. 20545
U.S. Atomic Energy Commission Atten: M.J. Whitman, Asst.Dir,
Division of Reactor Development for Program Analysis (1)
and Technology
Washington, D.C. 20545 U.S. Atomic Energy Conmission
Atten: Milton Shaw (1) Division of Reactor Development
and Technology
U.S. Atomic Energy Commission Washington, D.C. 20545
Division of Reactor Development Atten: Paul A. Halpine (1)
and Technology
Special Projects Branch U.S. Atomic Energy Conmission
Washington, D.C. 20545 Division of Reactor Development
Atten: R.F. Sweek (1) and Technology
Washington, D.C. 20545
U.S. Atomic Energy Commission Atten: G.W. Wensch, Chief
Division of Reactor Development Liquid Metals Projects, Br.(l)
and Technology
Washington^ D.C. 20545 U.S. Atomic Energy Conmission
Atten: J.W. Crawford, Asst.Dir. Division of Reactor Development
for Plant Engineering (1) and Technology
Reports Branch
U.S. Atomic Energy Commission Washington, D.C. 20545 (1)
Division of Reactor Development
and Technology U.S. Atomic Energy Conmission
Components Branch Headquarters Library
Washington, D.C, 20545 Washington, D.C- 20545 (1)
Atten: B.T. Resnick (1)
U.S. Atomic Energy Conmission
U.S. Atomic Energy Commission New York Operations Office
Division of Reactor Development Reports Librarian
and Technology 376 Hudson Street
Washington, D.C. 20545 New York, N.Y. 10014 (1)
Atten: E.E.Sinclair, Asst.Dir.
for Reactor Technology (1) U.S. Atomic Energy Commission
Brookhaven Office
Patents Office
Upton, N.Y. 11973 (1)
Distribution List (cont'd) Page 2
Contract No. AT(30-l)-3363

Brookhaven National Laboratory- Southwest Research Institute


Upton, New York 11973 8500 Culebra Road
Atten: D.O.E. Dwyer^ Chemical San Antonio, Texas 78206
Engineering Div. (1) Atten: Dr. R.A. Burton (1)

U.S. Atomic Energy Commission Chief of Naval Research


Division of Technical Information Department of the Navy
Extension Washington, D.C. 20360
P.O. Box 62 Atten: Code 438 (S.W.Doroff) (1)
Oak Ridge, Tennessee (3) 37830 Code 463 (Cmdr.R. Bodamer) (1)
(plus a completed AEC Form 426)
Head, Bearings and Seals Branch
Oak Ridge National Laboratory U.S. Naval Engineering Experiment Station
P.O. Box X Annapolis, Maryland 21402
Oak Ridge, Tennessee 37830 Atten: Code 851 (Watt V. Smith) (1)
Atten: A.P. Fraas, Rector Div.(l)
Commander
ArgOnne National Laboratory Aeronautical Systems Division
LMFBR Program Office of the Air Force Systems Command
9700 South Cass Avenue Wright-Patterson Air Force Base, Ohio 45433
Argonne, Illinois 60439 (1) Atten: ASRCNL-2, J.L. Morris (1)

Pacific Northwest Laboratory National Aeronautics & Space Administration


Fast Flux Test Facility Lewis Research Center
P.O. Box 999 21000 Brookpark Road
Richland, Washington 99352 Cleveland, Ohio 44135
Atten: E.R. Astley, Project Atten: Edmund E. Bisson, Chief
Manager, (1) Lubrication 6e Wear Branch (1)

U.S. Atomic Energy Commission Bearing and Lubricant Center


Space Nuclear Systems Division General Engineering Laboratory
Washington, D.C. 20545 General Electric Company
Atten: H.B. Finger, Dir. (1) 1 River Road
Schenectady, New York 12309
U.S. Atomic Energy Commission Atten: G.R. Fox, Manager (1)
Space Nuclear Systems Division
Washington, D.C. 20545 Franklin Institute
Atten: E.G. Johnson, Chief Laboratories for Research and Development
Facilities Branch (1) Philadelphia, Penna. 19103
Atten: Prof. D.D. Fuller (1)
Philco Corporation
Aeronutronic Division Columbia University
Newport Beach, California 92660 Department of Mechanical Engineering
Atten: H.D. Linhardt (1) New York, N.Y.
Atten: Prof. V. C a s t e l l i (1)
Prof. H.G. E l r o d , J r . (1)
Distribution List (cont'd) Page 3
Contract No. AT(30-1)-3363

Rocketdyne Division
Nucleonics Subdivision
6633 Canoga Avenue
Canoga Park, California 91304
Atten: Robert S. Siegler (1)

North American Aviation, Inc.


Atomics International Division
P.O. Box 309
Canoga Park, California 91305
Atten: R.W. Dickinson (1)


ADDITIONAL DISTRIBUTION LIST FOR REPORT NYO 3363-6

A.Warren Adam D.L. Deibel


Group Engineer Assist. Department Manager
Sundstrand Aviation TRW Incorporated
2421 Eleventh Street 23555 Euclid Avenue
Rockford, Illinois 61101 (1) Cleveland, Ohio 44117 (1)

G. Ambler Robert W. Dickinson


Department of the Navy Director, Sodium Component Operations
NAVSEC Atomics International
Washington, D.C. 20360 (1) P.O. Box 309
Canoga Park, California 91304 (1)
William J. Anderson
Chief, Bearings Branch Professor R. DiPrima
National Aeronautics & Space Admin. Mechanical Engineering Dept.
Lewis Research Center Rensselaer Polytechnic Institute
21000 Brookpark Road Troy, New York (1)
Cleveland, Ohio 44135 (1)
K. A. Domeisen
E.B. Arwas Pratt & Whitney Aircraft
Assistant to the Director-Technologies 400 Main Street
Mechanical Technology Inc. East Hartford, Conn. 06108 (1)
968 Albany-Shaker Road
Latham, New York 12110 (1) Dr. D. E. Dougherty
Senior Fluid Mechanics Engineer
R. Babbington Mechanical Technology Inc.
Space Nuclear Propulsion 968 Albany-Shaker Road
U.S. Atomic Energy Commission Latham,New York 12110 (1)
Washington, D.C. 20545 (1)
Mr. James Dray
R.E. Ball NAVSEC 6644
Assistant Chief Engineer Naval Ship Systems Command
Byron Jackson Pumps Inc. Department of the Mavy
P.O. Box 2017 Terminal Annex Washington, D.C. 20360 (1)
Los Angeles, California 90054 (1)
James H. Dunn
K. W. Benn NASA
Garrett Corporation Lewis Research Laboratory
402 South 36th Street 21000 Brookpark Road
Phoenix, Arizona (1) Cleveland, Ohio 44135 (1)

James E. Bocock Emil Eber


Nuclear Engineer Manager, Components Dept. SNAP-8 Div.
Navy Department Aerojet-General Corporation
Naval Facilities Engineering Command P.O. Box 296
Washington, D.C. 20390 (1) Azusa, California 91703 (1)

Dr. James Coyne R.C. Elwell


Bell Laboratories Atomic Motor Business Section
Mountain Avenue General Electric Company
Murray Hill, N. J. (1) 1 River Road, 50-360
Schenectady, New York 12301 (1)
Cmdr. F.L. Crump
Department of the Navy, ONR William G. Fibbi
Code 463 Engineer
Washington, D.C. 20360 (1) TRW Incorporated
23555 Euclid Ave.,
Richard Cundiff Cleveland, Ohio 44117 (1)
Development Engineer
Mechanical Technology Inc. A. P. Fraas
968 Albany-Shaker Road Assoc. Director, Reactor Division
Latham, New York 12110 (1) Oak Ridge National Laboratories
Bldg. 9102, Y-12
Otto Decker Oak Ridge, Tennessee 37830 (1)
Manager, Friction & Lubrication
Franklin Institute Research Lab. Nicholas Grossman
20th & Race Streets Chief, Special Technology Branch
Philadelphia, Penna. 19103 (1) U.S. Atomic Energy Commission
Div. of Reactor Development & Technol
Washington, D.C. 20545 (1)
R. Griswold S.F. Murray
Stein Seal Corporation Chemist
1920 Indiana Ave., Mechanical Technology Inc.
Philadelphia, Penna. (1) 968 Albany-Shaker Road
Latham, New York 12110 (1)
Robert D. Gruntz
Senior Project Engineer H.E. Nichols, Manager
AiResearch Manufacturing Co. of Arizona Turboinachinery Components
402 South 36th Street General Electric Company
Phoenix, Arizona 85034 (1) Missile and Space Division
Cincinnati, Ohio 45215 (1)
James Howard
Sr. Research Specialist Donald R. Nixon
Atomics International Design Engineer
P.O. Box 309 Westinghouse Atomic Equipment Div.
Canoga Park, California 91304 (1) Cheswick Ave.,
Cheswick, Penna. (1)
Edmund J. Jarski
Research Mechanical Engineer F.K. Orcutt
U.S. Navy Marine Engineering Lab. Research Engineer
Annapolis, Maryland 21402 (1) Mechanical Technology Inc.
968 Albany-Shaker Road
J.A. Kelly, Jr. Latham, iSIew York 12110 (1)
Manager, Marketing
Mechanical Technology Inc. Dr. C.H.T. Pan
968 Albany-Shaker Road Manager, Research
Latham, New York 12110 (1) Mechanical Technology Inc.
968 Albany-Shaker Road
R. Kasuba Latham, New York 12110
Principal Engineer
TRW Incorporated M.E. Peterson
23555 Euclid Ave. Manager, Engineering
Cleveland, Ohio 44117 (1) Atomic Motor Business Section
General Electric Company
E. King 1 River Road, 50-360
M.S.A. Research Corporation Schenectady, New York 12301 (1)
Evans City, Penna. (1)
D.E. Reardon
Robert E. Less ley SNAP Project Manager (Act.)
Senior Engineer SNAP Project Office
Aerojet-General Corporation U.S. Atomic Energy Commission
1100 Hollyvale Ave. P.O. Box 591
Azusa,California, (1) Canoga Park, California 91305 (1)

S. Luchter Cmdr. T. Robinson


Manager, Rankine Cycle Machinery Surface & Amphibious Program 463
Mechanical Technology Inc. ONR
968 Albany-Shaker Road M a i n Navy B l d g .
Latham, New York 12110 W a s h i n g t o n , D.C. 20360 (1)

Lawrence P. Ludwig R.D. Rossbach


Head, Seals Section Fluid & Thermodynamics Design
NASA Missile & Space Division
Lewis Research Center General Electric Company
21000 Brookpark Road Cincinnati, Ohio 45215 (1)
Cleveland, Ohio 44135
E. Schnetzer
J.C. McHugh Manager, Development Engineering
Project Engineer Space Power & Propulsion Section
Research & Development Center General Electric Company
General Electric Company 37-1020 Cincinnati, Ohio
P.O. Box 8
Schenectady, N.Y. (1)

Clarence E. Miller
Reactor Engineer
Division of Reactor Develop. & Tech.
Components Branch, U.S.Atomic Energy Coram.
Washington, D.C. 20545
Robert Spies J.R. Wetch
Supervisor Atomics International
Rocketdyne P.O. Box 309
6633 Canoga Ave., Canoga Park, California 91304 (1)
Canoga Park, California 91304 (1)
J. Yampolsky
Professor W.K. Stair Staff Member
Dept. of Mechanical & Aerospace Enginee General Atomic
The University of Tennessee P.O. Box 608
Knoxvtlle, Tennessee (1) San Diego, California (1)

Dr. Beno Sternllcht H. Young


Technical Director Oak Ridge National Laboratories
Mechanical Technology Inc. P.O. Box Y,
968 Albany-Shaker Road Oak Ridge, Tennesse 37830 (1)
Latham, Ne^ York 12110 (i)
Dr. N. Zuber
David Terrts Research & Development Center
Research Engineer RNDC
Atomics International Bldg. 37 - Rm. 609
P.O. Box 309 1 River Road
Canoga Park, California 91304 (1) Schenectady, N.Y. 12301 (1)

G. C. Thurston Dr. A.P. Villasor, Jr.


General Atomic Westinghouse Corporation
P.O. Box 608 Cheswick Ave.,
San Diego, California (1) Cheswick, Penna. (1)

Dr. J.H, Vohr A.D. Tonelli


Research Engineer Branch Chief Secondary Power Branch
Mechanical Technology Inc. Advance Biotechnology Dept.
968 Albany-Shaker Road Douglas Aircraft Go. Inc.
Latham, New Yorl 12110 (1) 3000 Ocean Park Blvd.
Santa Monica, California 90406 (1)
W. D. Waldron
Project Engineer Daniel Willens
Mechanical Technology Inc. Director, Systems Design Subdivision
968 Albany-Shaker Road Aerospace Corporation
Latham, New York 12110 (1) P.O. Box 95085
Los Angeles, California 90045 (1)
Horst Weinhold
Engineer Dudley D. Fuller
TRW Incorporated Professor
23555 Euclid Ave., Columbia University
Cleveland, Ohio 44117 (1) Dept. Mechanical Engineering
New York, N.Y. 10027 (1)
Dr. D.F. Wllcock
Manager R.J. Denington
Bearings, Seals & Lubrication Section Head
Mechanical Technology Inc. MSA
968 Albany-Shaker Rd. Lewis Research Center
Latham, New York 12110 (1) Cleveland, Ohio 44135 (1)

C.R. Winder John G- Krisllas


Principal Engineer Associate Head
tm Incorporated Power & Life Support Dept.
23555 Euclid Ave., Applied Mechanics Division
Clewland, Ohio 44117 (1) Aerospace Corporation
P.O. Box 95085
L. Wilson Los Angeles, California 90045 (1)
Oak Ridge National Laboratories
P.O. Box Y, Bldg. 9201-3
Oak lidge,Tennessee 37830 (1)

S-ar putea să vă placă și